nokogumbo 0.5 → 0.5.1

data/work/tokenizer.c

@@ -0,0 +1,2978 @@
+// Copyright 2010 Google Inc. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+// Author: jdtang@google.com (Jonathan Tang)
+//
+// Coding conventions specific to this file:
+//
+// 1. Functions that fill in a token should be named emit_*, and should be
+// followed immediately by a return from the tokenizer (true if no error
+// occurred, false if an error occurred).  Sometimes the emit functions
+// themselves return a boolean so that they can be combined with the return
+// statement; in this case, they should match this convention.
+// 2. Functions that shuffle data from temporaries to final API structures
+// should be named finish_*, and be called just before the tokenizer exits the
+// state that accumulates the temporary.
+// 3. All internal data structures should be kept in an initialized state from
+// tokenizer creation onwards, ready to accept input.  When a buffer's flushed
+// and reset, it should be deallocated and immediately reinitialized.
+// 4. Make sure there are appropriate break statements following each state.
+// 5. Assertions on the state of the temporary and tag buffers are usually a
+// good idea, and should go at the entry point of each state when added.
+// 6. Statement order within states goes:
+//    1. Add parse errors, if appropriate.
+//    2. Call finish_* functions to build up tag state.
+//    2. Switch to new state.  Set _reconsume flag if appropriate.
+//    3. Perform any other temporary buffer manipulation.
+//    4. Emit tokens
+//    5. Return/break.
+// This order ensures that we can verify that every emit is followed by a
+// return, ensures that the correct state is recorded with any parse errors, and
+// prevents parse error position from being messed up by possible mark/resets in
+// temporary buffer manipulation.
+
+
+#include "tokenizer.h"
+
+#include <assert.h>
+#include <stdbool.h>
+#include <string.h>
+
+#include "attribute.h"
+#include "char_ref.h"
+#include "error.h"
+#include "gumbo.h"
+#include "parser.h"
+#include "string_buffer.h"
+#include "string_piece.h"
+#include "token_type.h"
+#include "tokenizer_states.h"
+#include "utf8.h"
+#include "util.h"
+#include "vector.h"
+
+// Compared against _script_data_buffer to determine if we're in double-escaped
+// script mode.
+const GumboStringPiece kScriptTag = { "script", 6 };
+
+// An enum for the return value of each individual state.
+typedef enum {
+  RETURN_ERROR,         // Return false (error) from the tokenizer.
+  RETURN_SUCCESS,       // Return true (success) from the tokenizer.
+  NEXT_CHAR             // Proceed to the next character and continue lexing.
+} StateResult;
+
+// This is a struct containing state necessary to build up a tag token,
+// character by character.
+typedef struct _GumboTagState {
+  // A buffer to accumulate characters for various GumboStringPiece fields.
+  GumboStringBuffer _buffer;
+
+  // A pointer to the start of the original text corresponding to the contents
+  // of the buffer.
+  const char* _original_text;
+
+  // The current tag enum, computed once the tag name state has finished so that
+  // the buffer can be re-used for building up attributes.
+  GumboTag _tag;
+
+  // The starting location of the text in the buffer.
+  GumboSourcePosition _start_pos;
+
+  // The current list of attributes.  This is copied (and ownership of its data
+  // transferred) to the GumboStartTag token upon completion of the tag.  New
+  // attributes are added as soon as their attribute name state is complete, and
+  // values are filled in by operating on _attributes.data[attributes.length-1].
+  GumboVector /* GumboAttribute */ _attributes;
+
+  // If true, the next attribute value to be finished should be dropped.  This
+  // happens if a duplicate attribute name is encountered - we want to consume
+  // the attribute value, but shouldn't overwrite the existing value.
+  bool _drop_next_attr_value;
+
+  // The state that caused the tokenizer to switch into a character reference in
+  // attribute value state.  This is used to set the additional allowed
+  // character, and is switched back to on completion.  Initialized as the
+  // tokenizer enters the character reference state.
+  GumboTokenizerEnum _attr_value_state;
+
+  // The last start tag to have been emitted by the tokenizer.  This is
+  // necessary to check for appropriate end tags.
+  GumboTag _last_start_tag;
+
+  // If true, then this is a start tag.  If false, it's an end tag.  This is
+  // necessary to generate the appropriate token type at tag-closing time.
+  bool _is_start_tag;
+
+  // If true, then this tag is "self-closing" and doesn't have an end tag.
+  bool _is_self_closing;
+} GumboTagState;
+
+// This is the main tokenizer state struct, containing all state used by in
+// tokenizing the input stream.
+typedef struct _GumboTokenizerState {
+  // The current lexer state.  Starts in GUMBO_LEX_DATA.
+  GumboTokenizerEnum _state;
+
+  // A flag indicating whether the current input character needs to reconsumed
+  // in another state, or whether the next input character should be read for
+  // the next iteration of the state loop.  This is set when the spec reads
+  // "Reconsume the current input character in..."
+  bool _reconsume_current_input;
+
+  // A flag indicating whether the current node is a foreign element.  This is
+  // set by gumbo_tokenizer_set_is_current_node_foreign and checked in the
+  // markup declaration state.
+  bool _is_current_node_foreign;
+
+  // Certain states (notably character references) may emit two character tokens
+  // at once, but the contract for lex() fills in only one token at a time.  The
+  // extra character is buffered here, and then this is checked on entry to
+  // lex().  If a character is stored here, it's immediately emitted and control
+  // returns from the lexer.  kGumboNoChar is used to represent 'no character
+  // stored.'
+  //
+  // Note that characters emitted through this mechanism will have their source
+  // position marked as the character under the mark, i.e. multiple characters
+  // may be emitted with the same position.  This is desirable for character
+  // references, but unsuitable for many other cases.  Use the _temporary_buffer
+  // mechanism if the buffered characters must have their original positions in
+  // the document.
+  int _buffered_emit_char;
+
+  // A temporary buffer to accumulate characters, as described by the "temporary
+  // buffer" phrase in the tokenizer spec.  We use this in a somewhat unorthodox
+  // way: we record the specific character to go into the buffer, which may
+  // sometimes be a lowercased version of the actual input character.  However,
+  // we *also* use utf8iterator_mark() to record the position at tag start.
+  // When we start flushing the temporary buffer, we set _temporary_buffer_emit
+  // to the start of it, and then increment it for each call to the tokenizer.
+  // We also call utf8iterator_reset(), and utf8iterator_next() through the
+  // input stream, so that tokens emitted by emit_char have the correct position
+  // and original text.
+  GumboStringBuffer _temporary_buffer;
+
+  // The current cursor position we're emitting from within
+  // _temporary_buffer.data.  NULL whenever we're not flushing the buffer.
+  const char* _temporary_buffer_emit;
+
+  // The temporary buffer is also used by the spec to check whether we should
+  // enter the script data double escaped state, but we can't use the same
+  // buffer for both because we have to flush out "<s" as emits while still
+  // maintaining the context that will eventually become "script".  This is a
+  // separate buffer that's used in place of the temporary buffer for states
+  // that may enter the script data double escape start state.
+  GumboStringBuffer _script_data_buffer;
+
+  // Pointer to the beginning of the current token in the original buffer; used
+  // to record the original text.
+  const char* _token_start;
+
+  // GumboSourcePosition recording the source location of the start of the
+  // current token.
+  GumboSourcePosition _token_start_pos;
+
+  // Current tag state.
+  GumboTagState _tag_state;
+
+  // Doctype state.  We use the temporary buffer to accumulate characters (it's
+  // not used for anything else in the doctype states), and then freshly
+  // allocate the strings in the doctype token, then copy it over on emit.
+  GumboTokenDocType _doc_type_state;
+
+  // The UTF8Iterator over the tokenizer input.
+  Utf8Iterator _input;
+} GumboTokenizerState;
+
+// Adds an ERR_UNEXPECTED_CODE_POINT parse error to the parser's error struct.
+static void add_parse_error(GumboParser* parser, GumboErrorType type) {
+  GumboError* error = gumbo_add_error(parser);
+  if (!error) {
+    return;
+  }
+  GumboTokenizerState* tokenizer = parser->_tokenizer_state;
+  utf8iterator_get_position(&tokenizer->_input, &error->position);
+  error->original_text = utf8iterator_get_char_pointer(&tokenizer->_input);
+  error->type = type;
+  error->v.tokenizer.codepoint = utf8iterator_current(&tokenizer->_input);
+  switch (tokenizer->_state) {
+    case GUMBO_LEX_DATA:
+      error->v.tokenizer.state = GUMBO_ERR_TOKENIZER_DATA;
+      break;
+    case GUMBO_LEX_CHAR_REF_IN_DATA:
+    case GUMBO_LEX_CHAR_REF_IN_RCDATA:
+    case GUMBO_LEX_CHAR_REF_IN_ATTR_VALUE:
+      error->v.tokenizer.state = GUMBO_ERR_TOKENIZER_CHAR_REF;
+      break;
+    case GUMBO_LEX_RCDATA:
+    case GUMBO_LEX_RCDATA_LT:
+    case GUMBO_LEX_RCDATA_END_TAG_OPEN:
+    case GUMBO_LEX_RCDATA_END_TAG_NAME:
+      error->v.tokenizer.state = GUMBO_ERR_TOKENIZER_RCDATA;
+      break;
+    case GUMBO_LEX_RAWTEXT:
+    case GUMBO_LEX_RAWTEXT_LT:
+    case GUMBO_LEX_RAWTEXT_END_TAG_OPEN:
+    case GUMBO_LEX_RAWTEXT_END_TAG_NAME:
+      error->v.tokenizer.state = GUMBO_ERR_TOKENIZER_RAWTEXT;
+      break;
+    case GUMBO_LEX_PLAINTEXT:
+      error->v.tokenizer.state = GUMBO_ERR_TOKENIZER_PLAINTEXT;
+      break;
+    case GUMBO_LEX_SCRIPT:
+    case GUMBO_LEX_SCRIPT_LT:
+    case GUMBO_LEX_SCRIPT_END_TAG_OPEN:
+    case GUMBO_LEX_SCRIPT_END_TAG_NAME:
+    case GUMBO_LEX_SCRIPT_ESCAPED_START:
+    case GUMBO_LEX_SCRIPT_ESCAPED_START_DASH:
+    case GUMBO_LEX_SCRIPT_ESCAPED:
+    case GUMBO_LEX_SCRIPT_ESCAPED_DASH:
+    case GUMBO_LEX_SCRIPT_ESCAPED_DASH_DASH:
+    case GUMBO_LEX_SCRIPT_ESCAPED_LT:
+    case GUMBO_LEX_SCRIPT_ESCAPED_END_TAG_OPEN:
+    case GUMBO_LEX_SCRIPT_ESCAPED_END_TAG_NAME:
+    case GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED_START:
+    case GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED:
+    case GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED_DASH:
+    case GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED_DASH_DASH:
+    case GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED_LT:
+    case GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED_END:
+      error->v.tokenizer.state = GUMBO_ERR_TOKENIZER_SCRIPT;
+      break;
+    case GUMBO_LEX_TAG_OPEN:
+    case GUMBO_LEX_END_TAG_OPEN:
+    case GUMBO_LEX_TAG_NAME:
+    case GUMBO_LEX_BEFORE_ATTR_NAME:
+      error->v.tokenizer.state = GUMBO_ERR_TOKENIZER_TAG;
+      break;
+    case GUMBO_LEX_SELF_CLOSING_START_TAG:
+      error->v.tokenizer.state = GUMBO_ERR_TOKENIZER_SELF_CLOSING_TAG;
+      break;
+    case GUMBO_LEX_ATTR_NAME:
+    case GUMBO_LEX_AFTER_ATTR_NAME:
+    case GUMBO_LEX_BEFORE_ATTR_VALUE:
+      error->v.tokenizer.state = GUMBO_ERR_TOKENIZER_ATTR_NAME;
+      break;
+    case GUMBO_LEX_ATTR_VALUE_DOUBLE_QUOTED:
+    case GUMBO_LEX_ATTR_VALUE_SINGLE_QUOTED:
+    case GUMBO_LEX_ATTR_VALUE_UNQUOTED:
+    case GUMBO_LEX_AFTER_ATTR_VALUE_QUOTED:
+      error->v.tokenizer.state = GUMBO_ERR_TOKENIZER_ATTR_VALUE;
+      break;
+    case GUMBO_LEX_BOGUS_COMMENT:
+    case GUMBO_LEX_COMMENT_START:
+    case GUMBO_LEX_COMMENT_START_DASH:
+    case GUMBO_LEX_COMMENT:
+    case GUMBO_LEX_COMMENT_END_DASH:
+    case GUMBO_LEX_COMMENT_END:
+    case GUMBO_LEX_COMMENT_END_BANG:
+      error->v.tokenizer.state = GUMBO_ERR_TOKENIZER_COMMENT;
+      break;
+    case GUMBO_LEX_MARKUP_DECLARATION:
+    case GUMBO_LEX_DOCTYPE:
+    case GUMBO_LEX_BEFORE_DOCTYPE_NAME:
+    case GUMBO_LEX_DOCTYPE_NAME:
+    case GUMBO_LEX_AFTER_DOCTYPE_NAME:
+    case GUMBO_LEX_AFTER_DOCTYPE_PUBLIC_KEYWORD:
+    case GUMBO_LEX_BEFORE_DOCTYPE_PUBLIC_ID:
+    case GUMBO_LEX_DOCTYPE_PUBLIC_ID_DOUBLE_QUOTED:
+    case GUMBO_LEX_DOCTYPE_PUBLIC_ID_SINGLE_QUOTED:
+    case GUMBO_LEX_AFTER_DOCTYPE_PUBLIC_ID:
+    case GUMBO_LEX_BETWEEN_DOCTYPE_PUBLIC_SYSTEM_ID:
+    case GUMBO_LEX_AFTER_DOCTYPE_SYSTEM_KEYWORD:
+    case GUMBO_LEX_BEFORE_DOCTYPE_SYSTEM_ID:
+    case GUMBO_LEX_DOCTYPE_SYSTEM_ID_DOUBLE_QUOTED:
+    case GUMBO_LEX_DOCTYPE_SYSTEM_ID_SINGLE_QUOTED:
+    case GUMBO_LEX_AFTER_DOCTYPE_SYSTEM_ID:
+    case GUMBO_LEX_BOGUS_DOCTYPE:
+      error->v.tokenizer.state = GUMBO_ERR_TOKENIZER_DOCTYPE;
+      break;
+    case GUMBO_LEX_CDATA:
+      error->v.tokenizer.state = GUMBO_ERR_TOKENIZER_CDATA;
+      break;
+  }
+}
+
+static bool is_alpha(int c) {
+  // We don't use ISO C isupper/islower functions here because they
+  // depend upon the program's locale, while the behavior of the HTML5 spec is
+  // independent of which locale the program is run in.
+  return (c >= 'A' && c <= 'Z') || (c >= 'a' && c <= 'z');
+}
+
+static int ensure_lowercase(int c) {
+  return c >= 'A' && c <= 'Z' ? c + 0x20 : c;
+}
+
+static GumboTokenType get_char_token_type(int c) {
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\r':
+    case '\f':
+    case ' ':
+      return GUMBO_TOKEN_WHITESPACE;
+    case 0:
+      gumbo_debug("Emitted null byte.\n");
+      return GUMBO_TOKEN_NULL;
+    case -1:
+      return GUMBO_TOKEN_EOF;
+    default:
+      return GUMBO_TOKEN_CHARACTER;
+  }
+}
+
+// Starts recording characters in the temporary buffer.
+// Because this needs to reset the utf8iterator_mark to the beginning of the
+// text that will eventually be emitted, it needs to be called a couple of
+// states before the spec says "Set the temporary buffer to the empty string".
+// In general, this should be called whenever there's a transition to a
+// "less-than sign state".  The initial < and possibly / then need to be
+// appended to the temporary buffer, their presence needs to be accounted for in
+// states that compare the temporary buffer against a literal value, and
+// spec stanzas that say "emit a < and / character token along with a character
+// token for each character in the temporary buffer" need to be adjusted to
+// account for the presence of the < and / inside the temporary buffer.
+static void clear_temporary_buffer(GumboParser* parser) {
+  GumboTokenizerState* tokenizer = parser->_tokenizer_state;
+  assert(!tokenizer->_temporary_buffer_emit);
+  utf8iterator_mark(&tokenizer->_input);
+  gumbo_string_buffer_destroy(parser, &tokenizer->_temporary_buffer);
+  gumbo_string_buffer_init(parser, &tokenizer->_temporary_buffer);
+  // The temporary buffer and script data buffer are the same object in the
+  // spec, so the script data buffer should be cleared as well.
+  gumbo_string_buffer_destroy(parser, &tokenizer->_script_data_buffer);
+  gumbo_string_buffer_init(parser, &tokenizer->_script_data_buffer);
+}
+
+// Appends a codepoint to the temporary buffer.
+static void append_char_to_temporary_buffer(
+    GumboParser* parser, int codepoint) {
+  gumbo_string_buffer_append_codepoint(
+      parser, codepoint, &parser->_tokenizer_state->_temporary_buffer);
+}
+
+// Checks to see if the temporary buffer equals a certain string.
+// Make sure this remains side-effect free; it's used in assertions.
+#ifndef NDEBUG
+static bool temporary_buffer_equals(
+    GumboParser* parser, const char* text) {
+  GumboStringBuffer* buffer = &parser->_tokenizer_state->_temporary_buffer;
+  // TODO(jdtang): See if the extra strlen is a performance problem, and replace
+  // it with an explicit sizeof(literal) if necessary.  I don't think it will
+  // be, as this is only used in a couple of rare states.
+  int text_len = strlen(text);
+  return text_len == buffer->length &&
+      memcmp(buffer->data, text, text_len) == 0;
+}
+#endif
+
+static void doc_type_state_init(GumboParser* parser) {
+  GumboTokenDocType* doc_type_state =
+      &parser->_tokenizer_state->_doc_type_state;
+  // We initialize these to NULL here so that we don't end up leaking memory if
+  // we never see a doctype token.  When we do see a doctype token, we reset
+  // them to a freshly-allocated empty string so that we can present a uniform
+  // interface to client code and not make them check for null.  Ownership is
+  // transferred to the doctype token when it's emitted.
+  doc_type_state->name = NULL;
+  doc_type_state->public_identifier = NULL;
+  doc_type_state->system_identifier = NULL;
+  doc_type_state->force_quirks = false;
+  doc_type_state->has_public_identifier = false;
+  doc_type_state->has_system_identifier = false;
+}
+
+// Sets the token original_text and position to the current iterator position.
+// This is necessary because [CDATA[ sections may include text that is ignored
+// by the tokenizer.
+static void reset_token_start_point(GumboTokenizerState* tokenizer) {
+  tokenizer->_token_start = utf8iterator_get_char_pointer(&tokenizer->_input);
+  utf8iterator_get_position(&tokenizer->_input, &tokenizer->_token_start_pos);
+}
+
+// Sets the tag buffer original text and start point to the current iterator
+// position.  This is necessary because attribute names & values may have
+// whitespace preceeding them, and so we can't assume that the actual token
+// starting point was the end of the last tag buffer usage.
+static void reset_tag_buffer_start_point(GumboParser* parser) {
+  GumboTokenizerState* tokenizer = parser->_tokenizer_state;
+  GumboTagState* tag_state = &tokenizer->_tag_state;
+
+  utf8iterator_get_position(&tokenizer->_input, &tag_state->_start_pos);
+  tag_state->_original_text = utf8iterator_get_char_pointer(&tokenizer->_input);
+}
+
+// Moves the temporary buffer contents over to the specified output string,
+// and clears the temporary buffer.
+static void finish_temporary_buffer(GumboParser* parser, const char** output) {
+  GumboTokenizerState* tokenizer = parser->_tokenizer_state;
+  *output =
+      gumbo_string_buffer_to_string(parser, &tokenizer->_temporary_buffer);
+  clear_temporary_buffer(parser);
+}
+
+// Advances the iterator past the end of the token, and then fills in the
+// relevant position fields.  It's assumed that after every emit, the tokenizer
+// will immediately return (letting the tree-construction stage read the filled
+// in Token).  Thus, it's safe to advance the input stream here, since it will
+// bypass the advance at the bottom of the state machine loop.
+//
+// Since this advances the iterator and resets the current input, make sure to
+// call it after you've recorded any other data you need for the token.
+static void finish_token(GumboParser* parser, GumboToken* token) {
+  GumboTokenizerState* tokenizer = parser->_tokenizer_state;
+  if (!tokenizer->_reconsume_current_input) {
+    utf8iterator_next(&tokenizer->_input);
+  }
+
+  token->position = tokenizer->_token_start_pos;
+  token->original_text.data = tokenizer->_token_start;
+  reset_token_start_point(tokenizer);
+  token->original_text.length =
+      tokenizer->_token_start - token->original_text.data;
+  if (token->original_text.data[token->original_text.length - 1] == '\r') {
+    // The UTF8 iterator will ignore carriage returns in the input stream, which
+    // means that the next token may start one past a \r character.  The pointer
+    // arithmetic above results in that \r being appended to the original text
+    // of the preceding token, so we have to adjust its length here to chop the
+    // \r off.
+    --token->original_text.length;
+  }
+}
+
+// Records the doctype public ID, assumed to be in the temporary buffer.
+// Convenience method that also sets has_public_identifier to true.
+static void finish_doctype_public_id(GumboParser* parser) {
+  GumboTokenDocType* doc_type_state =
+      &parser->_tokenizer_state->_doc_type_state;
+  gumbo_parser_deallocate(parser, (void*) doc_type_state->public_identifier);
+  finish_temporary_buffer(parser, &doc_type_state->public_identifier);
+  doc_type_state->has_public_identifier = true;
+}
+
+// Records the doctype system ID, assumed to be in the temporary buffer.
+// Convenience method that also sets has_system_identifier to true.
+static void finish_doctype_system_id(GumboParser* parser) {
+  GumboTokenDocType* doc_type_state =
+      &parser->_tokenizer_state->_doc_type_state;
+  gumbo_parser_deallocate(parser, (void*) doc_type_state->system_identifier);
+  finish_temporary_buffer(parser, &doc_type_state->system_identifier);
+  doc_type_state->has_system_identifier = true;
+}
+
+// Writes a single specified character to the output token.
+static void emit_char(GumboParser* parser, int c, GumboToken* output) {
+  output->type = get_char_token_type(c);
+  output->v.character = c;
+  finish_token(parser, output);
+}
+
+// Writes a replacement character token and records a parse error.
+// Always returns RETURN_ERROR, per gumbo_lex return value.
+static StateResult emit_replacement_char(
+    GumboParser* parser, GumboToken* output) {
+  // In all cases, this is because of a null byte in the input stream.
+  add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+  emit_char(parser, kUtf8ReplacementChar, output);
+  return RETURN_ERROR;
+}
+
+// Writes an EOF character token.  Always returns RETURN_SUCCESS.
+static StateResult emit_eof(GumboParser* parser, GumboToken* output) {
+  emit_char(parser, -1, output);
+  return RETURN_SUCCESS;
+}
+
+// Writes the current input character out as a character token.
+// Always returns RETURN_SUCCESS.
+static bool emit_current_char(GumboParser* parser, GumboToken* output) {
+  emit_char(
+      parser, utf8iterator_current(&parser->_tokenizer_state->_input), output);
+  return RETURN_SUCCESS;
+}
+
+// Writes out a doctype token, copying it from the tokenizer state.
+static void emit_doctype(GumboParser* parser, GumboToken* output) {
+  output->type = GUMBO_TOKEN_DOCTYPE;
+  output->v.doc_type = parser->_tokenizer_state->_doc_type_state;
+  finish_token(parser, output);
+  doc_type_state_init(parser);
+}
+
+// Debug-only function that explicitly sets the attribute vector data to NULL so
+// it can be asserted on tag creation, verifying that there are no memory leaks.
+static void mark_tag_state_as_empty(GumboTagState* tag_state) {
+#ifndef NDEBUG
+  tag_state->_attributes = kGumboEmptyVector;
+#endif
+}
+
+// Writes out the current tag as a start or end tag token.
+// Always returns RETURN_SUCCESS.
+static StateResult emit_current_tag(GumboParser* parser, GumboToken* output) {
+  GumboTagState* tag_state = &parser->_tokenizer_state->_tag_state;
+  if (tag_state->_is_start_tag) {
+    output->type = GUMBO_TOKEN_START_TAG;
+    output->v.start_tag.tag = tag_state->_tag;
+    output->v.start_tag.attributes = tag_state->_attributes;
+    output->v.start_tag.is_self_closing = tag_state->_is_self_closing;
+    tag_state->_last_start_tag = tag_state->_tag;
+    mark_tag_state_as_empty(tag_state);
+    gumbo_debug("Emitted start tag %s.\n",
+               gumbo_normalized_tagname(tag_state->_tag));
+  } else {
+    output->type = GUMBO_TOKEN_END_TAG;
+    output->v.end_tag = tag_state->_tag;
+    // In end tags, ownership of the attributes vector is not transferred to the
+    // token, but it's still initialized as normal, so it must be manually
+    // deallocated.  There may also be attributes to destroy, in certain broken
+    // cases like </div</th> (the "th" is an attribute there).
+    for (int i = 0; i < tag_state->_attributes.length; ++i) {
+      gumbo_destroy_attribute(parser, tag_state->_attributes.data[i]);
+    }
+    gumbo_parser_deallocate(parser, tag_state->_attributes.data);
+    mark_tag_state_as_empty(tag_state);
+    gumbo_debug("Emitted end tag %s.\n",
+               gumbo_normalized_tagname(tag_state->_tag));
+  }
+  gumbo_string_buffer_destroy(parser, &tag_state->_buffer);
+  finish_token(parser, output);
+  gumbo_debug("Original text = %.*s.\n", output->original_text.length, output->original_text.data);
+  assert(output->original_text.length >= 2);
+  assert(output->original_text.data[0] == '<');
+  assert(output->original_text.data[output->original_text.length - 1] == '>');
+  return RETURN_SUCCESS;
+}
+
+// In some states, we speculatively start a tag, but don't know whether it'll be
+// emitted as tag token or as a series of character tokens until we finish it.
+// We need to abandon the tag we'd started & free its memory in that case to
+// avoid a memory leak.
+static void abandon_current_tag(GumboParser* parser) {
+  GumboTagState* tag_state = &parser->_tokenizer_state->_tag_state;
+  for (int i = 0; i < tag_state->_attributes.length; ++i) {
+    gumbo_destroy_attribute(parser, tag_state->_attributes.data[i]);
+  }
+  gumbo_parser_deallocate(parser, tag_state->_attributes.data);
+  mark_tag_state_as_empty(tag_state);
+  gumbo_string_buffer_destroy(parser, &tag_state->_buffer);
+  gumbo_debug("Abandoning current tag.\n");
+}
+
+// Wraps the consume_char_ref function to handle its output and make the
+// appropriate TokenizerState modifications.  Returns RETURN_ERROR if a parse
+// error occurred, RETURN_SUCCESS otherwise.
+static StateResult emit_char_ref(
+    GumboParser* parser, int additional_allowed_char,
+    bool is_in_attribute, GumboToken* output) {
+  GumboTokenizerState* tokenizer = parser->_tokenizer_state;
+  OneOrTwoCodepoints char_ref;
+  bool status = consume_char_ref(
+      parser, &tokenizer->_input, additional_allowed_char, false, &char_ref);
+  if (char_ref.first != kGumboNoChar) {
+    // consume_char_ref ends with the iterator pointing at the next character,
+    // so we need to be sure not advance it again before reading the next token.
+    tokenizer->_reconsume_current_input = true;
+    emit_char(parser, char_ref.first, output);
+    tokenizer->_buffered_emit_char = char_ref.second;
+  } else {
+    emit_char(parser, '&', output);
+  }
+  return status ? RETURN_SUCCESS : RETURN_ERROR;
+}
+
+// Emits a comment token.  Comments use the temporary buffer to accumulate their
+// data, and then it's copied over and released to the 'text' field of the
+// GumboToken union.  Always returns RETURN_SUCCESS.
+static StateResult emit_comment(GumboParser* parser, GumboToken* output) {
+  output->type = GUMBO_TOKEN_COMMENT;
+  finish_temporary_buffer(parser, &output->v.text);
+  finish_token(parser, output);
+  return RETURN_SUCCESS;
+}
+
+// Checks to see we should be flushing accumulated characters in the temporary
+// buffer, and fills the output token with the next output character if so.
+// Returns true if a character has been emitted and the tokenizer should
+// immediately return, false if we're at the end of the temporary buffer and
+// should resume normal operation.
+static bool maybe_emit_from_temporary_buffer(
+    GumboParser* parser, GumboToken* output) {
+  GumboTokenizerState* tokenizer = parser->_tokenizer_state;
+  const char* c = tokenizer->_temporary_buffer_emit;
+  GumboStringBuffer* buffer = &tokenizer->_temporary_buffer;
+
+  if (!c || c >= buffer->data + buffer->length) {
+    tokenizer->_temporary_buffer_emit = NULL;
+    return false;
+  }
+
+  assert(*c == utf8iterator_current(&tokenizer->_input));
+  // emit_char also advances the input stream.  We need to do some juggling of
+  // the _reconsume_current_input flag to get the proper behavior when emitting
+  // previous tokens.  Basically, _reconsume_current_input should *never* be set
+  // when emitting anything from the temporary buffer, since those characters
+  // have already been advanced past.  However, it should be preserved so that
+  // when the *next* character is encountered again, the tokenizer knows not to
+  // advance past it.
+  bool saved_reconsume_state = tokenizer->_reconsume_current_input;
+  tokenizer->_reconsume_current_input = false;
+  emit_char(parser, *c, output);
+  ++tokenizer->_temporary_buffer_emit;
+  tokenizer->_reconsume_current_input = saved_reconsume_state;
+  return true;
+}
+
+// Sets up the tokenizer to begin flushing the temporary buffer.
+// This resets the input iterator stream to the start of the last tag, sets up
+// _temporary_buffer_emit, and then (if the temporary buffer is non-empty) emits
+// the first character in it.  It returns true if a character was emitted, false
+// otherwise.
+static bool emit_temporary_buffer(
+    GumboParser* parser, GumboToken* output) {
+  GumboTokenizerState* tokenizer = parser->_tokenizer_state;
+  assert(tokenizer->_temporary_buffer.data);
+  utf8iterator_reset(&tokenizer->_input);
+  tokenizer->_temporary_buffer_emit = tokenizer->_temporary_buffer.data;
+  return maybe_emit_from_temporary_buffer(parser, output);
+}
+
+// Appends a codepoint to the current tag buffer.  If
+// reinitilize_position_on_first is set, this also initializes the tag buffer
+// start point; the only time you would *not* want to pass true for this
+// parameter is if you want the original_text to include character (like an
+// opening quote) that doesn't appear in the value.
+static void append_char_to_tag_buffer(GumboParser* parser, int codepoint,
+                                      bool reinitilize_position_on_first) {
+  GumboStringBuffer* buffer = &parser->_tokenizer_state->_tag_state._buffer;
+  if (buffer->length == 0 && reinitilize_position_on_first) {
+    reset_tag_buffer_start_point(parser);
+  }
+  gumbo_string_buffer_append_codepoint(parser, codepoint, buffer);
+}
+
+// (Re-)initialize the tag buffer.  This also resets the original_text pointer
+// and _start_pos field to point to the current position.
+static void initialize_tag_buffer(GumboParser* parser) {
+  GumboTokenizerState* tokenizer = parser->_tokenizer_state;
+  GumboTagState* tag_state = &tokenizer->_tag_state;
+
+  gumbo_string_buffer_init(parser, &tag_state->_buffer);
+  reset_tag_buffer_start_point(parser);
+}
+
+// Initializes the tag_state to start a new tag, keeping track of the opening
+// positions and original text.  Takes a boolean indicating whether this is a
+// start or end tag.
+static void start_new_tag(GumboParser* parser, bool is_start_tag) {
+  GumboTokenizerState* tokenizer = parser->_tokenizer_state;
+  GumboTagState* tag_state = &tokenizer->_tag_state;
+  int c = utf8iterator_current(&tokenizer->_input);
+  assert(is_alpha(c));
+  c = ensure_lowercase(c);
+  assert(is_alpha(c));
+
+  initialize_tag_buffer(parser);
+  gumbo_string_buffer_append_codepoint(parser, c, &tag_state->_buffer);
+
+  assert(tag_state->_attributes.data == NULL);
+  gumbo_vector_init(parser, 4, &tag_state->_attributes);
+  tag_state->_drop_next_attr_value = false;
+  tag_state->_is_start_tag = is_start_tag;
+  tag_state->_is_self_closing = false;
+  gumbo_debug("Starting new tag.\n");
+}
+
+// Fills in the specified char* with the contents of the tag buffer.
+static void copy_over_tag_buffer(GumboParser* parser, const char** output) {
+  GumboTokenizerState* tokenizer = parser->_tokenizer_state;
+  GumboTagState* tag_state = &tokenizer->_tag_state;
+  *output = gumbo_string_buffer_to_string(parser, &tag_state->_buffer);
+}
+
+// Fills in:
+// * The original_text GumboStringPiece with the portion of the original
+// buffer that corresponds to the tag buffer.
+// * The start_pos GumboSourcePosition with the start position of the tag
+// buffer.
+// * The end_pos GumboSourcePosition with the current source position.
+static void copy_over_original_tag_text(
+    GumboParser* parser, GumboStringPiece* original_text,
+    GumboSourcePosition* start_pos, GumboSourcePosition* end_pos) {
+  GumboTokenizerState* tokenizer = parser->_tokenizer_state;
+  GumboTagState* tag_state = &tokenizer->_tag_state;
+
+  original_text->data = tag_state->_original_text;
+  original_text->length =
+      utf8iterator_get_char_pointer(&tokenizer->_input) -
+      tag_state->_original_text;
+  if (original_text->data[original_text->length - 1] == '\r') {
+    // Since \r is skipped by the UTF-8 iterator, it can sometimes end up
+    // appended to the end of original text even when it's really the first part
+    // of the next character.  If we detect this situation, shrink the length of
+    // the original text by 1 to remove the carriage return.
+    --original_text->length;
+  }
+  *start_pos = tag_state->_start_pos;
+  utf8iterator_get_position(&tokenizer->_input, end_pos);
+}
+
+// Releases and then re-initializes the tag buffer.
+static void reinitialize_tag_buffer(GumboParser* parser) {
+  gumbo_parser_deallocate(
+      parser, parser->_tokenizer_state->_tag_state._buffer.data);
+  initialize_tag_buffer(parser);
+}
+
+// Moves some data from the temporary buffer over the the tag-based fields in
+// TagState.
+static void finish_tag_name(GumboParser* parser) {
+  GumboTokenizerState* tokenizer = parser->_tokenizer_state;
+  GumboTagState* tag_state = &tokenizer->_tag_state;
+
+  const char* temp;
+  copy_over_tag_buffer(parser, &temp);
+  tag_state->_tag = gumbo_tag_enum(temp);
+  reinitialize_tag_buffer(parser);
+  gumbo_parser_deallocate(parser, (void*) temp);
+}
+
+// Adds an ERR_DUPLICATE_ATTR parse error to the parser's error struct.
+static void add_duplicate_attr_error(GumboParser* parser, const char* attr_name,
+                                     int original_index, int new_index) {
+  GumboError* error = gumbo_add_error(parser);
+  if (!error) {
+    return;
+  }
+  GumboTagState* tag_state = &parser->_tokenizer_state->_tag_state;
+  error->type = GUMBO_ERR_DUPLICATE_ATTR;
+  error->position = tag_state->_start_pos;
+  error->original_text = tag_state->_original_text;
+  error->v.duplicate_attr.original_index = original_index;
+  error->v.duplicate_attr.new_index = new_index;
+  copy_over_tag_buffer(parser, &error->v.duplicate_attr.name);
+  reinitialize_tag_buffer(parser);
+}
+
+// Creates a new attribute in the current tag, copying the current tag buffer to
+// the attribute's name.  The attribute's value starts out as the empty string
+// (following the "Boolean attributes" section of the spec) and is only
+// overwritten on finish_attribute_value().  If the attribute has already been
+// specified, the new attribute is dropped, a parse error is added, and the
+// function returns false.  Otherwise, this returns true.
+static bool finish_attribute_name(GumboParser* parser) {
+  GumboTokenizerState* tokenizer = parser->_tokenizer_state;
+  GumboTagState* tag_state = &tokenizer->_tag_state;
+  // May've been set by a previous attribute without a value; reset it here.
+  tag_state->_drop_next_attr_value = false;
+  assert(tag_state->_attributes.data);
+  assert(tag_state->_attributes.capacity);
+
+  GumboVector* /* GumboAttribute* */ attributes = &tag_state->_attributes;
+  for (int i = 0; i < attributes->length; ++i) {
+    GumboAttribute* attr = attributes->data[i];
+    if (strlen(attr->name) == tag_state->_buffer.length &&
+        memcmp(attr->name, tag_state->_buffer.data,
+               tag_state->_buffer.length) == 0) {
+      // Identical attribute; bail.
+      add_duplicate_attr_error(
+          parser, attr->name, i, attributes->length);
+      tag_state->_drop_next_attr_value = true;
+      return false;
+    }
+  }
+
+  GumboAttribute* attr = gumbo_parser_allocate(parser, sizeof(GumboAttribute));
+  attr->attr_namespace = GUMBO_ATTR_NAMESPACE_NONE;
+  copy_over_tag_buffer(parser, &attr->name);
+  copy_over_original_tag_text(parser, &attr->original_name,
+                              &attr->name_start, &attr->name_end);
+  attr->value = gumbo_copy_stringz(parser, "");
+  copy_over_original_tag_text(parser, &attr->original_value,
+                              &attr->name_start, &attr->name_end);
+  gumbo_vector_add(parser, attr, attributes);
+  reinitialize_tag_buffer(parser);
+  return true;
+}
+
+// Finishes an attribute value.  This sets the value of the most recently added
+// attribute to the current contents of the tag buffer.
+static void finish_attribute_value(GumboParser* parser) {
+  GumboTagState* tag_state = &parser->_tokenizer_state->_tag_state;
+  if (tag_state->_drop_next_attr_value) {
+    // Duplicate attribute name detected in an earlier state, so we have to
+    // ignore the value.
+    tag_state->_drop_next_attr_value = false;
+    return;
+  }
+
+  GumboAttribute* attr =
+      tag_state->_attributes.data[tag_state->_attributes.length - 1];
+  gumbo_parser_deallocate(parser, (void*) attr->value);
+  copy_over_tag_buffer(parser, &attr->value);
+  copy_over_original_tag_text(parser, &attr->original_value,
+                              &attr->value_start, &attr->value_end);
+  reinitialize_tag_buffer(parser);
+}
+
+// Returns true if the current end tag matches the last start tag emitted.
+static bool is_appropriate_end_tag(GumboParser* parser) {
+  GumboTagState* tag_state = &parser->_tokenizer_state->_tag_state;
+  assert(!tag_state->_is_start_tag);
+  // Null terminate the current string buffer, so it can be passed to
+  // gumbo_tag_enum, but don't increment the length in case we need to dump the
+  // buffer as character tokens.
+  gumbo_string_buffer_append_codepoint(parser, '\0', &tag_state->_buffer);
+  --tag_state->_buffer.length;
+  return tag_state->_last_start_tag != GUMBO_TAG_LAST &&
+      tag_state->_last_start_tag == gumbo_tag_enum(tag_state->_buffer.data);
+}
+
+void gumbo_tokenizer_state_init(
+    GumboParser* parser, const char* text, size_t text_length) {
+  GumboTokenizerState* tokenizer =
+      gumbo_parser_allocate(parser, sizeof(GumboTokenizerState));
+  parser->_tokenizer_state = tokenizer;
+  gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+  tokenizer->_reconsume_current_input = false;
+  tokenizer->_is_current_node_foreign = false;
+  tokenizer->_tag_state._last_start_tag = GUMBO_TAG_LAST;
+
+  tokenizer->_buffered_emit_char = kGumboNoChar;
+  gumbo_string_buffer_init(parser, &tokenizer->_temporary_buffer);
+  tokenizer->_temporary_buffer_emit = NULL;
+
+  mark_tag_state_as_empty(&tokenizer->_tag_state);
+
+  gumbo_string_buffer_init(parser, &tokenizer->_script_data_buffer);
+  tokenizer->_token_start = text;
+  utf8iterator_init(parser, text, text_length, &tokenizer->_input);
+  utf8iterator_get_position(&tokenizer->_input, &tokenizer->_token_start_pos);
+  doc_type_state_init(parser);
+}
+
+void gumbo_tokenizer_state_destroy(GumboParser* parser) {
+  GumboTokenizerState* tokenizer = parser->_tokenizer_state;
+  assert(tokenizer->_doc_type_state.name == NULL);
+  assert(tokenizer->_doc_type_state.public_identifier == NULL);
+  assert(tokenizer->_doc_type_state.system_identifier == NULL);
+  gumbo_string_buffer_destroy(parser, &tokenizer->_temporary_buffer);
+  gumbo_string_buffer_destroy(parser, &tokenizer->_script_data_buffer);
+  gumbo_parser_deallocate(parser, tokenizer);
+}
+
+void gumbo_tokenizer_set_state(GumboParser* parser, GumboTokenizerEnum state) {
+  parser->_tokenizer_state->_state = state;
+}
+
+void gumbo_tokenizer_set_is_current_node_foreign(
+    GumboParser* parser, bool is_foreign) {
+  if (is_foreign != parser->_tokenizer_state->_is_current_node_foreign) {
+    gumbo_debug("Toggling is_current_node_foreign to %s.\n",
+                is_foreign ? "true" : "false");
+  }
+  parser->_tokenizer_state->_is_current_node_foreign = is_foreign;
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#data-state
+static StateResult handle_data_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '&':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_CHAR_REF_IN_DATA);
+      // The char_ref machinery expects to be on the & so it can mark that
+      // and return to it if the text isn't a char ref, so we need to
+      // reconsume it.
+      tokenizer->_reconsume_current_input = true;
+      return NEXT_CHAR;
+    case '<':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_TAG_OPEN);
+      clear_temporary_buffer(parser);
+      append_char_to_temporary_buffer(parser, '<');
+      return NEXT_CHAR;
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      emit_char(parser, c, output);
+      return RETURN_ERROR;
+    default:
+      return emit_current_char(parser, output);
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#character-reference-in-data-state
+static StateResult handle_char_ref_in_data_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+  return emit_char_ref(parser, ' ', false, output);
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#rcdata-state
+static StateResult handle_rcdata_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '&':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_CHAR_REF_IN_RCDATA);
+      tokenizer->_reconsume_current_input = true;
+      return NEXT_CHAR;
+    case '<':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_RCDATA_LT);
+      clear_temporary_buffer(parser);
+      append_char_to_temporary_buffer(parser, '<');
+      return NEXT_CHAR;
+    case '\0':
+      return emit_replacement_char(parser, output);
+    case -1:
+      return emit_eof(parser, output);
+    default:
+      return emit_current_char(parser, output);
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#character-reference-in-rcdata-state
+static StateResult handle_char_ref_in_rcdata_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  gumbo_tokenizer_set_state(parser, GUMBO_LEX_RCDATA);
+  return emit_char_ref(parser, ' ', false, output);
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#rawtext-state
+static StateResult handle_rawtext_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '<':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_RAWTEXT_LT);
+      clear_temporary_buffer(parser);
+      append_char_to_temporary_buffer(parser, '<');
+      return NEXT_CHAR;
+    case '\0':
+      return emit_replacement_char(parser, output);
+    case -1:
+      return emit_eof(parser, output);
+    default:
+      return emit_current_char(parser, output);
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#script-data-state
+static StateResult handle_script_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '<':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_LT);
+      clear_temporary_buffer(parser);
+      append_char_to_temporary_buffer(parser, '<');
+      return NEXT_CHAR;
+    case '\0':
+      return emit_replacement_char(parser, output);
+    case -1:
+      return emit_eof(parser, output);
+    default:
+      return emit_current_char(parser, output);
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#plaintext-state
+static StateResult handle_plaintext_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\0':
+      return emit_replacement_char(parser, output);
+    case -1:
+      return emit_eof(parser, output);
+    default:
+      return emit_current_char(parser, output);
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#tag-open-state
+static StateResult handle_tag_open_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  assert(temporary_buffer_equals(parser, "<"));
+  switch (c) {
+    case '!':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_MARKUP_DECLARATION);
+      clear_temporary_buffer(parser);
+      return NEXT_CHAR;
+    case '/':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_END_TAG_OPEN);
+      append_char_to_temporary_buffer(parser, '/');
+      return NEXT_CHAR;
+    case '?':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_BOGUS_COMMENT);
+      clear_temporary_buffer(parser);
+      append_char_to_temporary_buffer(parser, '?');
+      add_parse_error(parser, GUMBO_ERR_TAG_STARTS_WITH_QUESTION);
+      return NEXT_CHAR;
+    default:
+      if (is_alpha(c)) {
+        gumbo_tokenizer_set_state(parser, GUMBO_LEX_TAG_NAME);
+        start_new_tag(parser, true);
+        return NEXT_CHAR;
+      } else {
+        add_parse_error(parser, GUMBO_ERR_TAG_INVALID);
+        gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+        emit_temporary_buffer(parser, output);
+        return RETURN_ERROR;
+      }
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#end-tag-open-state
+static StateResult handle_end_tag_open_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  assert(temporary_buffer_equals(parser, "</"));
+  switch (c) {
+    case '>':
+      add_parse_error(parser, GUMBO_ERR_CLOSE_TAG_EMPTY);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      return NEXT_CHAR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_CLOSE_TAG_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      return emit_temporary_buffer(parser, output);
+    default:
+      if (is_alpha(c)) {
+        gumbo_tokenizer_set_state(parser, GUMBO_LEX_TAG_NAME);
+        start_new_tag(parser, false);
+      } else {
+        add_parse_error(parser, GUMBO_ERR_CLOSE_TAG_INVALID);
+        gumbo_tokenizer_set_state(parser, GUMBO_LEX_BOGUS_COMMENT);
+        clear_temporary_buffer(parser);
+        append_char_to_temporary_buffer(parser, c);
+      }
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#tag-name-state
+static StateResult handle_tag_name_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+      finish_tag_name(parser);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_BEFORE_ATTR_NAME);
+      return NEXT_CHAR;
+    case '/':
+      finish_tag_name(parser);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SELF_CLOSING_START_TAG);
+      return NEXT_CHAR;
+    case '>':
+      finish_tag_name(parser);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      return emit_current_tag(parser, output);
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      append_char_to_tag_buffer(parser, kUtf8ReplacementChar, true);
+      return NEXT_CHAR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_TAG_EOF);
+      abandon_current_tag(parser);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      return NEXT_CHAR;
+    default:
+      append_char_to_tag_buffer(parser, ensure_lowercase(c), true);
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#rcdata-less-than-sign-state
+static StateResult handle_rcdata_lt_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  assert(temporary_buffer_equals(parser, "<"));
+  if (c == '/') {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_RCDATA_END_TAG_OPEN);
+    append_char_to_temporary_buffer(parser, '/');
+    return NEXT_CHAR;
+  } else {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_RCDATA);
+    tokenizer->_reconsume_current_input = true;
+    return emit_temporary_buffer(parser, output);
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#rcdata-end-tag-open-state
+static StateResult handle_rcdata_end_tag_open_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  assert(temporary_buffer_equals(parser, "</"));
+  if (is_alpha(c)) {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_RCDATA_END_TAG_NAME);
+    start_new_tag(parser, false);
+    append_char_to_temporary_buffer(parser, c);
+    return NEXT_CHAR;
+  } else {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_RCDATA);
+    return emit_temporary_buffer(parser, output);
+  }
+  return true;
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#rcdata-end-tag-name-state
+static StateResult handle_rcdata_end_tag_name_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  assert(tokenizer->_temporary_buffer.length >= 2);
+  if (is_alpha(c)) {
+    append_char_to_tag_buffer(parser, ensure_lowercase(c), true);
+    append_char_to_temporary_buffer(parser, c);
+    return NEXT_CHAR;
+  } else if (is_appropriate_end_tag(parser)) {
+    switch (c) {
+      case '\t':
+      case '\n':
+      case '\f':
+      case ' ':
+        finish_tag_name(parser);
+        gumbo_tokenizer_set_state(parser, GUMBO_LEX_BEFORE_ATTR_NAME);
+        return NEXT_CHAR;
+      case '/':
+        finish_tag_name(parser);
+        gumbo_tokenizer_set_state(parser, GUMBO_LEX_SELF_CLOSING_START_TAG);
+        return NEXT_CHAR;
+      case '>':
+        finish_tag_name(parser);
+        gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+        return emit_current_tag(parser, output);
+    }
+  }
+  gumbo_tokenizer_set_state(parser, GUMBO_LEX_RCDATA);
+  abandon_current_tag(parser);
+  return emit_temporary_buffer(parser, output);
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#rawtext-less-than-sign-state
+static StateResult handle_rawtext_lt_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  assert(temporary_buffer_equals(parser, "<"));
+  if (c == '/') {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_RAWTEXT_END_TAG_OPEN);
+    append_char_to_temporary_buffer(parser, '/');
+    return NEXT_CHAR;
+  } else {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_RAWTEXT);
+    tokenizer->_reconsume_current_input = true;
+    return emit_temporary_buffer(parser, output);
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#rawtext-end-tag-open-state
+static StateResult handle_rawtext_end_tag_open_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  assert(temporary_buffer_equals(parser, "</"));
+  if (is_alpha(c)) {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_RAWTEXT_END_TAG_NAME);
+    start_new_tag(parser, false);
+    append_char_to_temporary_buffer(parser, c);
+    return NEXT_CHAR;
+  } else {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_RAWTEXT);
+    return emit_temporary_buffer(parser, output);
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#rawtext-end-tag-name-state
+static StateResult handle_rawtext_end_tag_name_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  assert(tokenizer->_temporary_buffer.length >= 2);
+  gumbo_debug("Last end tag: %*s\n", (int) tokenizer->_tag_state._buffer.length,
+             tokenizer->_tag_state._buffer.data);
+  if (is_alpha(c)) {
+    append_char_to_tag_buffer(parser, ensure_lowercase(c), true);
+    append_char_to_temporary_buffer(parser, c);
+    return NEXT_CHAR;
+  } else if (is_appropriate_end_tag(parser)) {
+    gumbo_debug("Is an appropriate end tag.\n");
+    switch (c) {
+      case '\t':
+      case '\n':
+      case '\f':
+      case ' ':
+        finish_tag_name(parser);
+        gumbo_tokenizer_set_state(parser, GUMBO_LEX_BEFORE_ATTR_NAME);
+        return NEXT_CHAR;
+      case '/':
+        finish_tag_name(parser);
+        gumbo_tokenizer_set_state(parser, GUMBO_LEX_SELF_CLOSING_START_TAG);
+        return NEXT_CHAR;
+      case '>':
+        finish_tag_name(parser);
+        gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+        return emit_current_tag(parser, output);
+    }
+  }
+  gumbo_tokenizer_set_state(parser, GUMBO_LEX_RAWTEXT);
+  abandon_current_tag(parser);
+  return emit_temporary_buffer(parser, output);
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#script-data-less-than-sign-state
+static StateResult handle_script_lt_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  assert(temporary_buffer_equals(parser, "<"));
+  if (c == '/') {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_END_TAG_OPEN);
+    append_char_to_temporary_buffer(parser, '/');
+    return NEXT_CHAR;
+  } else if (c == '!') {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_ESCAPED_START);
+    append_char_to_temporary_buffer(parser, '!');
+    return emit_temporary_buffer(parser, output);
+  } else {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT);
+    tokenizer->_reconsume_current_input = true;
+    return emit_temporary_buffer(parser, output);
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#script-data-end-tag-open-state
+static StateResult handle_script_end_tag_open_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  assert(temporary_buffer_equals(parser, "</"));
+  if (is_alpha(c)) {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_END_TAG_NAME);
+    start_new_tag(parser, false);
+    append_char_to_temporary_buffer(parser, c);
+    return NEXT_CHAR;
+  } else {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT);
+    return emit_temporary_buffer(parser, output);
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#script-data-end-tag-name-state
+static StateResult handle_script_end_tag_name_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  assert(tokenizer->_temporary_buffer.length >= 2);
+  if (is_alpha(c)) {
+    append_char_to_tag_buffer(parser, ensure_lowercase(c), true);
+    append_char_to_temporary_buffer(parser, c);
+    return NEXT_CHAR;
+  } else if (is_appropriate_end_tag(parser)) {
+    switch (c) {
+      case '\t':
+      case '\n':
+      case '\f':
+      case ' ':
+        finish_tag_name(parser);
+        gumbo_tokenizer_set_state(parser, GUMBO_LEX_BEFORE_ATTR_NAME);
+        return NEXT_CHAR;
+      case '/':
+        finish_tag_name(parser);
+        gumbo_tokenizer_set_state(parser, GUMBO_LEX_SELF_CLOSING_START_TAG);
+        return NEXT_CHAR;
+      case '>':
+        finish_tag_name(parser);
+        gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+        return emit_current_tag(parser, output);
+    }
+  }
+  gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT);
+  abandon_current_tag(parser);
+  return emit_temporary_buffer(parser, output);
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#script-data-escape-start-state
+static StateResult handle_script_escaped_start_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  if (c == '-') {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_ESCAPED_START_DASH);
+    return emit_current_char(parser, output);
+  } else {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT);
+    tokenizer->_reconsume_current_input = true;
+    return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#script-data-escape-start-dash-state
+static StateResult handle_script_escaped_start_dash_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  if (c == '-') {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_ESCAPED_DASH_DASH);
+    return emit_current_char(parser, output);
+  } else {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT);
+    tokenizer->_reconsume_current_input = true;
+    return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#script-data-escaped-state
+static StateResult handle_script_escaped_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '-':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_ESCAPED_DASH);
+      return emit_current_char(parser, output);
+    case '<':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_ESCAPED_LT);
+      clear_temporary_buffer(parser);
+      append_char_to_temporary_buffer(parser, c);
+      return NEXT_CHAR;
+    case '\0':
+      return emit_replacement_char(parser, output);
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_SCRIPT_EOF);
+      return emit_eof(parser, output);
+    default:
+      return emit_current_char(parser, output);
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#script-data-escaped-dash-state
+static StateResult handle_script_escaped_dash_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '-':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_ESCAPED_DASH_DASH);
+      return emit_current_char(parser, output);
+    case '<':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_ESCAPED_LT);
+      clear_temporary_buffer(parser);
+      append_char_to_temporary_buffer(parser, c);
+      return NEXT_CHAR;
+    case '\0':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_ESCAPED);
+      return emit_replacement_char(parser, output);
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_SCRIPT_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      return NEXT_CHAR;
+    default:
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_ESCAPED);
+      return emit_current_char(parser, output);
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#script-data-escaped-dash-dash-state
+static StateResult handle_script_escaped_dash_dash_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '-':
+      return emit_current_char(parser, output);
+    case '<':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_ESCAPED_LT);
+      clear_temporary_buffer(parser);
+      append_char_to_temporary_buffer(parser, c);
+      return NEXT_CHAR;
+    case '>':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT);
+      return emit_current_char(parser, output);
+    case '\0':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_ESCAPED);
+      return emit_replacement_char(parser, output);
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_SCRIPT_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      return NEXT_CHAR;
+    default:
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_ESCAPED);
+      return emit_current_char(parser, output);
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#script-data-escaped-less-than-sign-state
+static StateResult handle_script_escaped_lt_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  assert(temporary_buffer_equals(parser, "<"));
+  assert(!tokenizer->_script_data_buffer.length);
+  if (c == '/') {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_ESCAPED_END_TAG_OPEN);
+    append_char_to_temporary_buffer(parser, c);
+    return NEXT_CHAR;
+  } else if (is_alpha(c)) {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED_START);
+    append_char_to_temporary_buffer(parser, c);
+    gumbo_string_buffer_append_codepoint(
+        parser, ensure_lowercase(c), &tokenizer->_script_data_buffer);
+    return emit_temporary_buffer(parser, output);
+  } else {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_ESCAPED);
+    return emit_temporary_buffer(parser, output);
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#script-data-escaped-end-tag-open-state
+static StateResult handle_script_escaped_end_tag_open_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  assert(temporary_buffer_equals(parser, "</"));
+  if (is_alpha(c)) {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_ESCAPED_END_TAG_NAME);
+    start_new_tag(parser, false);
+    append_char_to_temporary_buffer(parser, c);
+    return NEXT_CHAR;
+  } else {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_ESCAPED);
+    return emit_temporary_buffer(parser, output);
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#script-data-escaped-end-tag-name-state
+static StateResult handle_script_escaped_end_tag_name_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  assert(tokenizer->_temporary_buffer.length >= 2);
+  if (is_alpha(c)) {
+    append_char_to_tag_buffer(parser, ensure_lowercase(c), true);
+    append_char_to_temporary_buffer(parser, c);
+    return NEXT_CHAR;
+  } else if (is_appropriate_end_tag(parser)) {
+    switch (c) {
+      case '\t':
+      case '\n':
+      case '\f':
+      case ' ':
+        finish_tag_name(parser);
+        gumbo_tokenizer_set_state(parser, GUMBO_LEX_BEFORE_ATTR_NAME);
+        return NEXT_CHAR;
+      case '/':
+        finish_tag_name(parser);
+        gumbo_tokenizer_set_state(parser, GUMBO_LEX_SELF_CLOSING_START_TAG);
+        return NEXT_CHAR;
+      case '>':
+        finish_tag_name(parser);
+        gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+        return emit_current_tag(parser, output);
+    }
+  }
+  gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_ESCAPED);
+  abandon_current_tag(parser);
+  return emit_temporary_buffer(parser, output);
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#script-data-double-escape-start-state
+static StateResult handle_script_double_escaped_start_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+    case '/':
+    case '>':
+      gumbo_tokenizer_set_state(parser, gumbo_string_equals(
+          &kScriptTag, (GumboStringPiece*) &tokenizer->_script_data_buffer)
+          ? GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED : GUMBO_LEX_SCRIPT_ESCAPED);
+      return emit_current_char(parser, output);
+    default:
+      if (is_alpha(c)) {
+        gumbo_string_buffer_append_codepoint(
+            parser, ensure_lowercase(c), &tokenizer->_script_data_buffer);
+        return emit_current_char(parser, output);
+      } else {
+        gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_ESCAPED);
+        tokenizer->_reconsume_current_input = true;
+        return NEXT_CHAR;
+      }
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#script-data-double-escaped-state
+static StateResult handle_script_double_escaped_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '-':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED_DASH);
+      return emit_current_char(parser, output);
+    case '<':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED_LT);
+      return emit_current_char(parser, output);
+    case '\0':
+      return emit_replacement_char(parser, output);
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_SCRIPT_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      return NEXT_CHAR;
+    default:
+      return emit_current_char(parser, output);
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#script-data-double-escaped-dash-state
+static StateResult handle_script_double_escaped_dash_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '-':
+      gumbo_tokenizer_set_state(
+          parser, GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED_DASH_DASH);
+      return emit_current_char(parser, output);
+    case '<':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED_LT);
+      return emit_current_char(parser, output);
+    case '\0':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED);
+      return emit_replacement_char(parser, output);
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_SCRIPT_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      return NEXT_CHAR;
+    default:
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED);
+      return emit_current_char(parser, output);
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#script-data-double-escaped-dash-dash-state
+static StateResult handle_script_double_escaped_dash_dash_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '-':
+      return emit_current_char(parser, output);
+    case '<':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED_LT);
+      return emit_current_char(parser, output);
+    case '>':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT);
+      return emit_current_char(parser, output);
+    case '\0':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED);
+      return emit_replacement_char(parser, output);
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_SCRIPT_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      return NEXT_CHAR;
+    default:
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED);
+      return emit_current_char(parser, output);
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#script-data-double-escaped-less-than-sign-state
+static StateResult handle_script_double_escaped_lt_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  if (c == '/') {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED_END);
+    gumbo_string_buffer_destroy(parser, &tokenizer->_script_data_buffer);
+    gumbo_string_buffer_init(parser, &tokenizer->_script_data_buffer);
+    return emit_current_char(parser, output);
+  } else {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED);
+    tokenizer->_reconsume_current_input = true;
+    return NEXT_CHAR;
+  }
+
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#script-data-double-escape-end-state
+static StateResult handle_script_double_escaped_end_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+    case '/':
+    case '>':
+      gumbo_tokenizer_set_state(parser, gumbo_string_equals(
+          &kScriptTag, (GumboStringPiece*) &tokenizer->_script_data_buffer)
+          ? GUMBO_LEX_SCRIPT_ESCAPED : GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED);
+      return emit_current_char(parser, output);
+    default:
+      if (is_alpha(c)) {
+        gumbo_string_buffer_append_codepoint(
+            parser, ensure_lowercase(c), &tokenizer->_script_data_buffer);
+        return emit_current_char(parser, output);
+      } else {
+        gumbo_tokenizer_set_state(parser, GUMBO_LEX_SCRIPT_DOUBLE_ESCAPED);
+        tokenizer->_reconsume_current_input = true;
+        return NEXT_CHAR;
+      }
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#before-attribute-name-state
+static StateResult handle_before_attr_name_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+      return NEXT_CHAR;
+    case '/':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SELF_CLOSING_START_TAG);
+      return NEXT_CHAR;
+    case '>':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      return emit_current_tag(parser, output);
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_ATTR_NAME);
+      append_char_to_temporary_buffer(parser, 0xfffd);
+      return NEXT_CHAR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_ATTR_NAME_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      abandon_current_tag(parser);
+      return NEXT_CHAR;
+    case '"':
+    case '\'':
+    case '<':
+    case '=':
+      add_parse_error(parser, GUMBO_ERR_ATTR_NAME_INVALID);
+      // Fall through.
+    default:
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_ATTR_NAME);
+      append_char_to_tag_buffer(parser, ensure_lowercase(c), true);
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#attribute-name-state
+static StateResult handle_attr_name_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+      finish_attribute_name(parser);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_AFTER_ATTR_NAME);
+      return NEXT_CHAR;
+    case '/':
+      finish_attribute_name(parser);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SELF_CLOSING_START_TAG);
+      return NEXT_CHAR;
+    case '=':
+      finish_attribute_name(parser);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_BEFORE_ATTR_VALUE);
+      return NEXT_CHAR;
+    case '>':
+      finish_attribute_name(parser);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      return emit_current_tag(parser, output);
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      append_char_to_tag_buffer(parser, kUtf8ReplacementChar, true);
+      return NEXT_CHAR;
+    case -1:
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      abandon_current_tag(parser);
+      add_parse_error(parser, GUMBO_ERR_ATTR_NAME_EOF);
+      return NEXT_CHAR;
+    case '"':
+    case '\'':
+    case '<':
+      add_parse_error(parser, GUMBO_ERR_ATTR_NAME_INVALID);
+      // Fall through.
+    default:
+      append_char_to_tag_buffer(parser, ensure_lowercase(c), true);
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#after-attribute-name-state
+static StateResult handle_after_attr_name_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+      return NEXT_CHAR;
+    case '/':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SELF_CLOSING_START_TAG);
+      return NEXT_CHAR;
+    case '=':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_BEFORE_ATTR_VALUE);
+      return NEXT_CHAR;
+    case '>':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      return emit_current_tag(parser, output);
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_ATTR_NAME);
+      append_char_to_temporary_buffer(parser, 0xfffd);
+      return NEXT_CHAR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_ATTR_NAME_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      abandon_current_tag(parser);
+      return NEXT_CHAR;
+    case '"':
+    case '\'':
+    case '<':
+      add_parse_error(parser, GUMBO_ERR_ATTR_NAME_INVALID);
+      // Fall through.
+    default:
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_ATTR_NAME);
+      append_char_to_tag_buffer(parser, ensure_lowercase(c), true);
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#before-attribute-value-state
+static StateResult handle_before_attr_value_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+      return NEXT_CHAR;
+    case '"':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_ATTR_VALUE_DOUBLE_QUOTED);
+      reset_tag_buffer_start_point(parser);
+      return NEXT_CHAR;
+    case '&':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_ATTR_VALUE_UNQUOTED);
+      tokenizer->_reconsume_current_input = true;
+      return NEXT_CHAR;
+    case '\'':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_ATTR_VALUE_SINGLE_QUOTED);
+      reset_tag_buffer_start_point(parser);
+      return NEXT_CHAR;
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_ATTR_VALUE_UNQUOTED);
+      append_char_to_tag_buffer(parser, kUtf8ReplacementChar, true);
+      return NEXT_CHAR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_ATTR_UNQUOTED_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      abandon_current_tag(parser);
+      tokenizer->_reconsume_current_input = true;
+      return NEXT_CHAR;
+    case '>':
+      add_parse_error(parser, GUMBO_ERR_ATTR_UNQUOTED_RIGHT_BRACKET);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      emit_current_tag(parser, output);
+      return RETURN_ERROR;
+    case '<':
+    case '=':
+    case '`':
+      add_parse_error(parser, GUMBO_ERR_ATTR_UNQUOTED_EQUALS);
+      // Fall through.
+    default:
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_ATTR_VALUE_UNQUOTED);
+      append_char_to_tag_buffer(parser, c, true);
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#attribute-value-double-quoted-state
+static StateResult handle_attr_value_double_quoted_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '"':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_AFTER_ATTR_VALUE_QUOTED);
+      return NEXT_CHAR;
+    case '&':
+      tokenizer->_tag_state._attr_value_state = tokenizer->_state;
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_CHAR_REF_IN_ATTR_VALUE);
+      tokenizer->_reconsume_current_input = true;
+      return NEXT_CHAR;
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      append_char_to_tag_buffer(parser, kUtf8ReplacementChar, false);
+      return NEXT_CHAR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_ATTR_DOUBLE_QUOTE_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      abandon_current_tag(parser);
+      tokenizer->_reconsume_current_input = true;
+      return NEXT_CHAR;
+    default:
+      append_char_to_tag_buffer(parser, c, false);
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#attribute-value-single-quoted-state
+static StateResult handle_attr_value_single_quoted_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\'':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_AFTER_ATTR_VALUE_QUOTED);
+      return NEXT_CHAR;
+    case '&':
+      tokenizer->_tag_state._attr_value_state = tokenizer->_state;
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_CHAR_REF_IN_ATTR_VALUE);
+      tokenizer->_reconsume_current_input = true;
+      return NEXT_CHAR;
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      append_char_to_tag_buffer(parser, kUtf8ReplacementChar, false);
+      return NEXT_CHAR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_ATTR_SINGLE_QUOTE_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      abandon_current_tag(parser);
+      tokenizer->_reconsume_current_input = true;
+      return NEXT_CHAR;
+    default:
+      append_char_to_tag_buffer(parser, c, false);
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#attribute-value-unquoted-state
+static StateResult handle_attr_value_unquoted_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_BEFORE_ATTR_NAME);
+      finish_attribute_value(parser);
+      return NEXT_CHAR;
+    case '&':
+      tokenizer->_tag_state._attr_value_state = tokenizer->_state;
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_CHAR_REF_IN_ATTR_VALUE);
+      tokenizer->_reconsume_current_input = true;
+      return NEXT_CHAR;
+    case '>':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      finish_attribute_value(parser);
+      return emit_current_tag(parser, output);
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      append_char_to_tag_buffer(parser, kUtf8ReplacementChar, true);
+      return NEXT_CHAR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_ATTR_UNQUOTED_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_reconsume_current_input = true;
+      abandon_current_tag(parser);
+      return NEXT_CHAR;
+    case '<':
+    case '=':
+    case '"':
+    case '\'':
+    case '`':
+      add_parse_error(parser, GUMBO_ERR_ATTR_UNQUOTED_EQUALS);
+      // Fall through.
+    default:
+      append_char_to_tag_buffer(parser, c, true);
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#character-reference-in-attribute-value-state
+static StateResult handle_char_ref_in_attr_value_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  OneOrTwoCodepoints char_ref;
+  int allowed_char;
+  bool is_unquoted = false;
+  switch (tokenizer->_tag_state._attr_value_state) {
+    case GUMBO_LEX_ATTR_VALUE_DOUBLE_QUOTED:
+      allowed_char = '"';
+      break;
+    case GUMBO_LEX_ATTR_VALUE_SINGLE_QUOTED:
+      allowed_char = '\'';
+      break;
+    case GUMBO_LEX_ATTR_VALUE_UNQUOTED:
+      allowed_char = '>';
+      is_unquoted = true;
+      break;
+    default:
+      // -Wmaybe-uninitialized is a little overzealous here, and doesn't
+      // get that the assert(0) means this codepath will never happen.
+      allowed_char = ' ';
+      assert(0);
+  }
+
+  // Ignore the status, since we don't have a convenient way of signalling that
+  // a parser error has occurred when the error occurs in the middle of a
+  // multi-state token.  We'd need a flag inside the TokenizerState to do this,
+  // but that's a low priority fix.
+  consume_char_ref(parser, &tokenizer->_input, allowed_char, true, &char_ref);
+  if (char_ref.first != kGumboNoChar) {
+    tokenizer->_reconsume_current_input = true;
+    append_char_to_tag_buffer(parser, char_ref.first, is_unquoted);
+    if (char_ref.second != kGumboNoChar) {
+      append_char_to_tag_buffer(parser, char_ref.second, is_unquoted);
+    }
+  } else {
+    append_char_to_tag_buffer(parser, '&', is_unquoted);
+  }
+  gumbo_tokenizer_set_state(parser, tokenizer->_tag_state._attr_value_state);
+  return NEXT_CHAR;
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#after-attribute-value-quoted-state
+static StateResult handle_after_attr_value_quoted_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  finish_attribute_value(parser);
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_BEFORE_ATTR_NAME);
+      return NEXT_CHAR;
+    case '/':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_SELF_CLOSING_START_TAG);
+      return NEXT_CHAR;
+    case '>':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      return emit_current_tag(parser, output);
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_ATTR_AFTER_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      abandon_current_tag(parser);
+      tokenizer->_reconsume_current_input = true;
+      return NEXT_CHAR;
+    default:
+      add_parse_error(parser, GUMBO_ERR_ATTR_AFTER_INVALID);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_BEFORE_ATTR_NAME);
+      tokenizer->_reconsume_current_input = true;
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#self-closing-start-tag-state
+static StateResult handle_self_closing_start_tag_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '>':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_tag_state._is_self_closing = true;
+      return emit_current_tag(parser, output);
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_SOLIDUS_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      abandon_current_tag(parser);
+      return NEXT_CHAR;
+    default:
+      add_parse_error(parser, GUMBO_ERR_SOLIDUS_INVALID);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_BEFORE_ATTR_NAME);
+      tokenizer->_reconsume_current_input = true;
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#bogus-comment-state
+static StateResult handle_bogus_comment_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  while (c != '>' && c != -1) {
+    if (c == '\0') {
+      c = 0xFFFD;
+    }
+    append_char_to_temporary_buffer(parser, c);
+    utf8iterator_next(&tokenizer->_input);
+    c = utf8iterator_current(&tokenizer->_input);
+  }
+  gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+  return emit_comment(parser, output);
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#markup-declaration-open-state
+static StateResult handle_markup_declaration_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  if (utf8iterator_maybe_consume_match(
+      &tokenizer->_input, "--", sizeof("--") - 1, true)) {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_COMMENT_START);
+    tokenizer->_reconsume_current_input = true;
+  } else if (utf8iterator_maybe_consume_match(
+      &tokenizer->_input, "DOCTYPE", sizeof("DOCTYPE") - 1, false)) {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_DOCTYPE);
+    tokenizer->_reconsume_current_input = true;
+    // If we get here, we know we'll eventually emit a doctype token, so now is
+    // the time to initialize the doctype strings.  (Not in doctype_state_init,
+    // since then they'll leak if ownership never gets transferred to the
+    // doctype token.
+    tokenizer->_doc_type_state.name = gumbo_copy_stringz(parser, "");
+    tokenizer->_doc_type_state.public_identifier =
+        gumbo_copy_stringz(parser, "");
+    tokenizer->_doc_type_state.system_identifier =
+        gumbo_copy_stringz(parser, "");
+  } else if (tokenizer->_is_current_node_foreign &&
+             utf8iterator_maybe_consume_match(
+                &tokenizer->_input, "[CDATA[", sizeof("[CDATA[") - 1, true)) {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_CDATA);
+    tokenizer->_reconsume_current_input = true;
+  } else {
+    add_parse_error(parser, GUMBO_ERR_DASHES_OR_DOCTYPE);
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_BOGUS_COMMENT);
+    tokenizer->_reconsume_current_input = true;
+    clear_temporary_buffer(parser);
+  }
+  return NEXT_CHAR;
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#comment-start-state
+static StateResult handle_comment_start_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '-':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_COMMENT_START_DASH);
+      return NEXT_CHAR;
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_COMMENT);
+      append_char_to_temporary_buffer(parser, kUtf8ReplacementChar);
+      return NEXT_CHAR;
+    case '>':
+      add_parse_error(parser, GUMBO_ERR_COMMENT_INVALID);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      emit_comment(parser, output);
+      return RETURN_ERROR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_COMMENT_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      emit_comment(parser, output);
+      return RETURN_ERROR;
+    default:
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_COMMENT);
+      append_char_to_temporary_buffer(parser, c);
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#comment-start-dash-state
+static StateResult handle_comment_start_dash_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '-':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_COMMENT_END);
+      return NEXT_CHAR;
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_COMMENT);
+      append_char_to_temporary_buffer(parser, '-');
+      append_char_to_temporary_buffer(parser, kUtf8ReplacementChar);
+      return NEXT_CHAR;
+    case '>':
+      add_parse_error(parser, GUMBO_ERR_COMMENT_INVALID);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      emit_comment(parser, output);
+      return RETURN_ERROR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_COMMENT_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      emit_comment(parser, output);
+      return RETURN_ERROR;
+    default:
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_COMMENT);
+      append_char_to_temporary_buffer(parser, '-');
+      append_char_to_temporary_buffer(parser, c);
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#comment-state
+static StateResult handle_comment_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '-':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_COMMENT_END_DASH);
+      return NEXT_CHAR;
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      append_char_to_temporary_buffer(parser, kUtf8ReplacementChar);
+      return NEXT_CHAR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_COMMENT_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      emit_comment(parser, output);
+      return RETURN_ERROR;
+    default:
+      append_char_to_temporary_buffer(parser, c);
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#comment-end-dash-state
+static StateResult handle_comment_end_dash_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '-':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_COMMENT_END);
+      return NEXT_CHAR;
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_COMMENT);
+      append_char_to_temporary_buffer(parser, '-');
+      append_char_to_temporary_buffer(parser, kUtf8ReplacementChar);
+      return NEXT_CHAR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_COMMENT_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      emit_comment(parser, output);
+      return RETURN_ERROR;
+    default:
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_COMMENT);
+      append_char_to_temporary_buffer(parser, '-');
+      append_char_to_temporary_buffer(parser, c);
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#comment-end-state
+static StateResult handle_comment_end_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '>':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      return emit_comment(parser, output);
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_COMMENT);
+      append_char_to_temporary_buffer(parser, '-');
+      append_char_to_temporary_buffer(parser, '-');
+      append_char_to_temporary_buffer(parser, kUtf8ReplacementChar);
+      return NEXT_CHAR;
+    case '!':
+      add_parse_error(parser, GUMBO_ERR_COMMENT_BANG_AFTER_DOUBLE_DASH);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_COMMENT_END_BANG);
+      return NEXT_CHAR;
+    case '-':
+      add_parse_error(parser, GUMBO_ERR_COMMENT_DASH_AFTER_DOUBLE_DASH);
+      append_char_to_temporary_buffer(parser, '-');
+      return NEXT_CHAR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      emit_comment(parser, output);
+      return RETURN_ERROR;
+    default:
+      add_parse_error(parser, GUMBO_ERR_COMMENT_INVALID);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_COMMENT);
+      append_char_to_temporary_buffer(parser, '-');
+      append_char_to_temporary_buffer(parser, '-');
+      append_char_to_temporary_buffer(parser, c);
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#comment-end-bang-state
+static StateResult handle_comment_end_bang_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '-':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_COMMENT_END_DASH);
+      append_char_to_temporary_buffer(parser, '-');
+      append_char_to_temporary_buffer(parser, '-');
+      append_char_to_temporary_buffer(parser, '!');
+      return NEXT_CHAR;
+    case '>':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      return emit_comment(parser, output);
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_COMMENT);
+      append_char_to_temporary_buffer(parser, '-');
+      append_char_to_temporary_buffer(parser, '-');
+      append_char_to_temporary_buffer(parser, '!');
+      append_char_to_temporary_buffer(parser, kUtf8ReplacementChar);
+      return NEXT_CHAR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_COMMENT_END_BANG_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      emit_comment(parser, output);
+      return RETURN_ERROR;
+    default:
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_COMMENT);
+      append_char_to_temporary_buffer(parser, '-');
+      append_char_to_temporary_buffer(parser, '-');
+      append_char_to_temporary_buffer(parser, '!');
+      append_char_to_temporary_buffer(parser, c);
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#doctype-state
+static StateResult handle_doctype_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  assert(!tokenizer->_temporary_buffer.length);
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_BEFORE_DOCTYPE_NAME);
+      return NEXT_CHAR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    default:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_SPACE);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_BEFORE_DOCTYPE_NAME);
+      tokenizer->_reconsume_current_input = true;
+      tokenizer->_doc_type_state.force_quirks = true;
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#before-doctype-name-state
+static StateResult handle_before_doctype_name_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+      return NEXT_CHAR;
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DOCTYPE_NAME);
+      tokenizer->_doc_type_state.force_quirks = true;
+      append_char_to_temporary_buffer(parser, kUtf8ReplacementChar);
+      return NEXT_CHAR;
+    case '>':
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_RIGHT_BRACKET);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    default:
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DOCTYPE_NAME);
+      tokenizer->_doc_type_state.force_quirks = false;
+      append_char_to_temporary_buffer(parser, ensure_lowercase(c));
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete5/tokenization.html#doctype-name-state
+static StateResult handle_doctype_name_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_AFTER_DOCTYPE_NAME);
+      gumbo_parser_deallocate(parser, (void*) tokenizer->_doc_type_state.name);
+      finish_temporary_buffer(
+          parser, &tokenizer->_doc_type_state.name);
+      return NEXT_CHAR;
+    case '>':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      gumbo_parser_deallocate(parser, (void*) tokenizer->_doc_type_state.name);
+      finish_temporary_buffer(
+          parser, &tokenizer->_doc_type_state.name);
+      emit_doctype(parser, output);
+      return RETURN_SUCCESS;
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      append_char_to_temporary_buffer(parser, kUtf8ReplacementChar);
+      return NEXT_CHAR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      gumbo_parser_deallocate(parser, (void*) tokenizer->_doc_type_state.name);
+      finish_temporary_buffer(
+          parser, &tokenizer->_doc_type_state.name);
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    default:
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DOCTYPE_NAME);
+      tokenizer->_doc_type_state.force_quirks = false;
+      append_char_to_temporary_buffer(parser, ensure_lowercase(c));
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#after-doctype-name-state
+static StateResult handle_after_doctype_name_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+      return NEXT_CHAR;
+    case '>':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      emit_doctype(parser, output);
+      return RETURN_SUCCESS;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    default:
+      if (utf8iterator_maybe_consume_match(
+          &tokenizer->_input, "PUBLIC", sizeof("PUBLIC") - 1, false)) {
+        gumbo_tokenizer_set_state(
+            parser, GUMBO_LEX_AFTER_DOCTYPE_PUBLIC_KEYWORD);
+        tokenizer->_reconsume_current_input = true;
+      } else if (utf8iterator_maybe_consume_match(
+          &tokenizer->_input, "SYSTEM", sizeof("SYSTEM") - 1, false)) {
+        gumbo_tokenizer_set_state(
+            parser, GUMBO_LEX_AFTER_DOCTYPE_SYSTEM_KEYWORD);
+        tokenizer->_reconsume_current_input = true;
+      } else {
+        add_parse_error(parser, GUMBO_ERR_DOCTYPE_SPACE_OR_RIGHT_BRACKET);
+        gumbo_tokenizer_set_state(parser, GUMBO_LEX_BOGUS_DOCTYPE);
+        tokenizer->_doc_type_state.force_quirks = true;
+      }
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#after-doctype-public-keyword-state
+static StateResult handle_after_doctype_public_keyword_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+      gumbo_tokenizer_set_state(
+          parser, GUMBO_LEX_BEFORE_DOCTYPE_PUBLIC_ID);
+      return NEXT_CHAR;
+    case '"':
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_INVALID);
+      assert(temporary_buffer_equals(parser, ""));
+      gumbo_tokenizer_set_state(
+          parser, GUMBO_LEX_DOCTYPE_PUBLIC_ID_DOUBLE_QUOTED);
+      return NEXT_CHAR;
+    case '\'':
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_INVALID);
+      assert(temporary_buffer_equals(parser, ""));
+      gumbo_tokenizer_set_state(
+          parser, GUMBO_LEX_DOCTYPE_PUBLIC_ID_SINGLE_QUOTED);
+      return NEXT_CHAR;
+    case '>':
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_RIGHT_BRACKET);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    default:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_INVALID);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_BOGUS_DOCTYPE);
+      tokenizer->_doc_type_state.force_quirks = true;
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#before-doctype-public-identifier-state
+static StateResult handle_before_doctype_public_id_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+      return NEXT_CHAR;
+    case '"':
+      assert(temporary_buffer_equals(parser, ""));
+      gumbo_tokenizer_set_state(
+          parser, GUMBO_LEX_DOCTYPE_PUBLIC_ID_DOUBLE_QUOTED);
+      return NEXT_CHAR;
+    case '\'':
+      assert(temporary_buffer_equals(parser, ""));
+      gumbo_tokenizer_set_state(
+          parser, GUMBO_LEX_DOCTYPE_PUBLIC_ID_SINGLE_QUOTED);
+      return NEXT_CHAR;
+    case '>':
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_END);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    default:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_INVALID);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_BOGUS_DOCTYPE);
+      tokenizer->_doc_type_state.force_quirks = true;
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#doctype-public-identifier-(double-quoted)-state
+static StateResult handle_doctype_public_id_double_quoted_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '"':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_AFTER_DOCTYPE_PUBLIC_ID);
+      finish_doctype_public_id(parser);
+      return NEXT_CHAR;
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      append_char_to_temporary_buffer(parser, kUtf8ReplacementChar);
+      return NEXT_CHAR;
+    case '>':
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_END);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      finish_doctype_public_id(parser);
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      finish_doctype_public_id(parser);
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    default:
+      append_char_to_temporary_buffer(parser, c);
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#doctype-public-identifier-(single-quoted)-state
+static StateResult handle_doctype_public_id_single_quoted_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\'':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_AFTER_DOCTYPE_PUBLIC_ID);
+      finish_doctype_public_id(parser);
+      return NEXT_CHAR;
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      append_char_to_temporary_buffer(parser, kUtf8ReplacementChar);
+      return NEXT_CHAR;
+    case '>':
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_END);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      finish_doctype_public_id(parser);
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      finish_doctype_public_id(parser);
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    default:
+      append_char_to_temporary_buffer(parser, c);
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#after-doctype-public-identifier-state
+static StateResult handle_after_doctype_public_id_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+      gumbo_tokenizer_set_state(
+          parser, GUMBO_LEX_BETWEEN_DOCTYPE_PUBLIC_SYSTEM_ID);
+      return NEXT_CHAR;
+    case '>':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      emit_doctype(parser, output);
+      return RETURN_SUCCESS;
+    case '"':
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_INVALID);
+      assert(temporary_buffer_equals(parser, ""));
+      gumbo_tokenizer_set_state(
+          parser, GUMBO_LEX_DOCTYPE_SYSTEM_ID_DOUBLE_QUOTED);
+      return NEXT_CHAR;
+    case '\'':
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_INVALID);
+      assert(temporary_buffer_equals(parser, ""));
+      gumbo_tokenizer_set_state(
+          parser, GUMBO_LEX_DOCTYPE_SYSTEM_ID_SINGLE_QUOTED);
+      return NEXT_CHAR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_reconsume_current_input = true;
+      tokenizer->_doc_type_state.force_quirks = true;
+      return NEXT_CHAR;
+    default:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_INVALID);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_BOGUS_DOCTYPE);
+      tokenizer->_doc_type_state.force_quirks = true;
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#between-doctype-public-and-system-identifiers-state
+static StateResult handle_between_doctype_public_system_id_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+      return NEXT_CHAR;
+    case '>':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      emit_doctype(parser, output);
+      return RETURN_SUCCESS;
+    case '"':
+      assert(temporary_buffer_equals(parser, ""));
+      gumbo_tokenizer_set_state(
+          parser, GUMBO_LEX_DOCTYPE_SYSTEM_ID_DOUBLE_QUOTED);
+      return NEXT_CHAR;
+    case '\'':
+      assert(temporary_buffer_equals(parser, ""));
+      gumbo_tokenizer_set_state(
+          parser, GUMBO_LEX_DOCTYPE_SYSTEM_ID_SINGLE_QUOTED);
+      return NEXT_CHAR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    default:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_INVALID);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_BOGUS_DOCTYPE);
+      tokenizer->_doc_type_state.force_quirks = true;
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#after-doctype-system-keyword-state
+static StateResult handle_after_doctype_system_keyword_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_BEFORE_DOCTYPE_SYSTEM_ID);
+      return NEXT_CHAR;
+    case '"':
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_INVALID);
+      assert(temporary_buffer_equals(parser, ""));
+      gumbo_tokenizer_set_state(
+          parser, GUMBO_LEX_DOCTYPE_SYSTEM_ID_DOUBLE_QUOTED);
+      return NEXT_CHAR;
+    case '\'':
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_INVALID);
+      assert(temporary_buffer_equals(parser, ""));
+      gumbo_tokenizer_set_state(
+          parser, GUMBO_LEX_DOCTYPE_SYSTEM_ID_SINGLE_QUOTED);
+      return NEXT_CHAR;
+    case '>':
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_END);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    default:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_INVALID);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_BOGUS_DOCTYPE);
+      tokenizer->_doc_type_state.force_quirks = true;
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#before-doctype-system-identifier-state
+static StateResult handle_before_doctype_system_id_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+      return NEXT_CHAR;
+    case '"':
+      assert(temporary_buffer_equals(parser, ""));
+      gumbo_tokenizer_set_state(
+          parser, GUMBO_LEX_DOCTYPE_SYSTEM_ID_DOUBLE_QUOTED);
+      return NEXT_CHAR;
+    case '\'':
+      assert(temporary_buffer_equals(parser, ""));
+      gumbo_tokenizer_set_state(
+          parser, GUMBO_LEX_DOCTYPE_SYSTEM_ID_SINGLE_QUOTED);
+      return NEXT_CHAR;
+    case '>':
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_END);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    default:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_INVALID);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_BOGUS_DOCTYPE);
+      tokenizer->_doc_type_state.force_quirks = true;
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#doctype-system-identifier-(double-quoted)-state
+static StateResult handle_doctype_system_id_double_quoted_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '"':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_AFTER_DOCTYPE_SYSTEM_ID);
+      finish_doctype_system_id(parser);
+      return NEXT_CHAR;
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      append_char_to_temporary_buffer(parser, kUtf8ReplacementChar);
+      return NEXT_CHAR;
+    case '>':
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_END);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      finish_doctype_system_id(parser);
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      finish_doctype_system_id(parser);
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    default:
+      append_char_to_temporary_buffer(parser, c);
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#doctype-system-identifier-(single-quoted)-state
+static StateResult handle_doctype_system_id_single_quoted_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\'':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_AFTER_DOCTYPE_SYSTEM_ID);
+      finish_doctype_system_id(parser);
+      return NEXT_CHAR;
+    case '\0':
+      add_parse_error(parser, GUMBO_ERR_UTF8_NULL);
+      append_char_to_temporary_buffer(parser, kUtf8ReplacementChar);
+      return NEXT_CHAR;
+    case '>':
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_END);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      finish_doctype_system_id(parser);
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_INVALID);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      finish_doctype_system_id(parser);
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    default:
+      append_char_to_temporary_buffer(parser, c);
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#after-doctype-system-identifier-state
+static StateResult handle_after_doctype_system_id_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  switch (c) {
+    case '\t':
+    case '\n':
+    case '\f':
+    case ' ':
+      return NEXT_CHAR;
+    case '>':
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      emit_doctype(parser, output);
+      return RETURN_SUCCESS;
+    case -1:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_EOF);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+      tokenizer->_doc_type_state.force_quirks = true;
+      emit_doctype(parser, output);
+      return RETURN_ERROR;
+    default:
+      add_parse_error(parser, GUMBO_ERR_DOCTYPE_INVALID);
+      gumbo_tokenizer_set_state(parser, GUMBO_LEX_BOGUS_DOCTYPE);
+      return NEXT_CHAR;
+  }
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#bogus-doctype-state
+static StateResult handle_bogus_doctype_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  if (c == '>' || c == -1) {
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+    emit_doctype(parser, output);
+    return RETURN_ERROR;
+  }
+  return NEXT_CHAR;
+}
+
+// http://www.whatwg.org/specs/web-apps/current-work/complete.html#cdata-section-state
+static StateResult handle_cdata_state(
+    GumboParser* parser, GumboTokenizerState* tokenizer,
+    int c, GumboToken* output) {
+  if (c == -1 || utf8iterator_maybe_consume_match(
+      &tokenizer->_input, "]]>", sizeof("]]>") - 1, true)) {
+    tokenizer->_reconsume_current_input = true;
+    reset_token_start_point(tokenizer);
+    gumbo_tokenizer_set_state(parser, GUMBO_LEX_DATA);
+    return NEXT_CHAR;
+  } else {
+    return emit_current_char(parser, output);
+  }
+}
+
+typedef StateResult (*GumboLexerStateFunction)(
+    GumboParser*, GumboTokenizerState*, int, GumboToken*);
+
+static GumboLexerStateFunction dispatch_table[] = {
+  handle_data_state,
+  handle_char_ref_in_data_state,
+  handle_rcdata_state,
+  handle_char_ref_in_rcdata_state,
+  handle_rawtext_state,
+  handle_script_state,
+  handle_plaintext_state,
+  handle_tag_open_state,
+  handle_end_tag_open_state,
+  handle_tag_name_state,
+  handle_rcdata_lt_state,
+  handle_rcdata_end_tag_open_state,
+  handle_rcdata_end_tag_name_state,
+  handle_rawtext_lt_state,
+  handle_rawtext_end_tag_open_state,
+  handle_rawtext_end_tag_name_state,
+  handle_script_lt_state,
+  handle_script_end_tag_open_state,
+  handle_script_end_tag_name_state,
+  handle_script_escaped_start_state,
+  handle_script_escaped_start_dash_state,
+  handle_script_escaped_state,
+  handle_script_escaped_dash_state,
+  handle_script_escaped_dash_dash_state,
+  handle_script_escaped_lt_state,
+  handle_script_escaped_end_tag_open_state,
+  handle_script_escaped_end_tag_name_state,
+  handle_script_double_escaped_start_state,
+  handle_script_double_escaped_state,
+  handle_script_double_escaped_dash_state,
+  handle_script_double_escaped_dash_dash_state,
+  handle_script_double_escaped_lt_state,
+  handle_script_double_escaped_end_state,
+  handle_before_attr_name_state,
+  handle_attr_name_state,
+  handle_after_attr_name_state,
+  handle_before_attr_value_state,
+  handle_attr_value_double_quoted_state,
+  handle_attr_value_single_quoted_state,
+  handle_attr_value_unquoted_state,
+  handle_char_ref_in_attr_value_state,
+  handle_after_attr_value_quoted_state,
+  handle_self_closing_start_tag_state,
+  handle_bogus_comment_state,
+  handle_markup_declaration_state,
+  handle_comment_start_state,
+  handle_comment_start_dash_state,
+  handle_comment_state,
+  handle_comment_end_dash_state,
+  handle_comment_end_state,
+  handle_comment_end_bang_state,
+  handle_doctype_state,
+  handle_before_doctype_name_state,
+  handle_doctype_name_state,
+  handle_after_doctype_name_state,
+  handle_after_doctype_public_keyword_state,
+  handle_before_doctype_public_id_state,
+  handle_doctype_public_id_double_quoted_state,
+  handle_doctype_public_id_single_quoted_state,
+  handle_after_doctype_public_id_state,
+  handle_between_doctype_public_system_id_state,
+  handle_after_doctype_system_keyword_state,
+  handle_before_doctype_system_id_state,
+  handle_doctype_system_id_double_quoted_state,
+  handle_doctype_system_id_single_quoted_state,
+  handle_after_doctype_system_id_state,
+  handle_bogus_doctype_state,
+  handle_cdata_state
+};
+
+bool gumbo_lex(GumboParser* parser, GumboToken* output) {
+  // Because of the spec requirements that...
+  //
+  // 1. Tokens be handled immediately by the parser upon emission.
+  // 2. Some states (eg. CDATA, or various error conditions) require the
+  // emission of multiple tokens in the same states.
+  // 3. The tokenizer often has to reconsume the same character in a different
+  // state.
+  //
+  // ...all state must be held in the GumboTokenizer struct instead of in local
+  // variables in this function.  That allows us to return from this method with
+  // a token, and then immediately jump back to the same state with the same
+  // input if we need to return a different token.  The various emit_* functions
+  // are responsible for changing state (eg. flushing the chardata buffer,
+  // reading the next input character) to avoid an infinite loop.
+  GumboTokenizerState* tokenizer = parser->_tokenizer_state;
+
+  if (tokenizer->_buffered_emit_char != kGumboNoChar) {
+    tokenizer->_reconsume_current_input = true;
+    emit_char(parser, tokenizer->_buffered_emit_char, output);
+    // And now that we've avoided advancing the input, make sure we set
+    // _reconsume_current_input back to false to make sure the *next* character
+    // isn't consumed twice.
+    tokenizer->_reconsume_current_input = false;
+    tokenizer->_buffered_emit_char = kGumboNoChar;
+    return true;
+  }
+
+  if (maybe_emit_from_temporary_buffer(parser, output)) {
+    return true;
+  }
+
+  while (1) {
+    assert(!tokenizer->_temporary_buffer_emit);
+    assert(tokenizer->_buffered_emit_char == kGumboNoChar);
+    int c = utf8iterator_current(&tokenizer->_input);
+    gumbo_debug("Lexing character '%c' in state %d.\n", c, tokenizer->_state);
+    StateResult result =
+        dispatch_table[tokenizer->_state](parser, tokenizer, c, output);
+    // We need to clear reconsume_current_input before returning to prevent
+    // certain infinite loop states.
+    bool should_advance = !tokenizer->_reconsume_current_input;
+    tokenizer->_reconsume_current_input = false;
+
+    if (result == RETURN_SUCCESS) {
+      return true;
+    } else if(result == RETURN_ERROR) {
+      return false;
+    }
+
+    if (should_advance) {
+      utf8iterator_next(&tokenizer->_input);
+    }
+  }
+}
+
+void gumbo_token_destroy(GumboParser* parser, GumboToken* token) {
+  if (!token) return;
+
+  switch (token->type) {
+    case GUMBO_TOKEN_DOCTYPE:
+      gumbo_parser_deallocate(parser, (void*) token->v.doc_type.name);
+      gumbo_parser_deallocate(
+          parser, (void*) token->v.doc_type.public_identifier);
+      gumbo_parser_deallocate(
+          parser, (void*) token->v.doc_type.system_identifier);
+      return;
+    case GUMBO_TOKEN_START_TAG:
+      for (int i = 0; i < token->v.start_tag.attributes.length; ++i) {
+        GumboAttribute* attr = token->v.start_tag.attributes.data[i];
+        if (attr) {
+          // May have been nulled out if this token was merged with another.
+          gumbo_destroy_attribute(parser, attr);
+        }
+      }
+      gumbo_parser_deallocate(
+          parser, (void*) token->v.start_tag.attributes.data);
+      return;
+    case GUMBO_TOKEN_COMMENT:
+      gumbo_parser_deallocate(parser, (void*) token->v.text);
+      return;
+    default:
+      return;
+  }
+}