nokogiri 1.10.10 → 1.12.1

data/ext/nokogiri/html4_document.c

@@ -0,0 +1,166 @@
+#include <nokogiri.h>
+
+VALUE cNokogiriHtml4Document ;
+
+static ID id_encoding_found;
+static ID id_to_s;
+
+/*
+ * call-seq:
+ *  new
+ *
+ * Create a new document
+ */
+static VALUE
+rb_html_document_s_new(int argc, VALUE *argv, VALUE klass)
+{
+  VALUE uri, external_id, rest, rb_doc;
+  htmlDocPtr doc;
+
+  rb_scan_args(argc, argv, "0*", &rest);
+  uri = rb_ary_entry(rest, (long)0);
+  external_id = rb_ary_entry(rest, (long)1);
+
+  doc = htmlNewDoc(
+          RTEST(uri) ? (const xmlChar *)StringValueCStr(uri) : NULL,
+          RTEST(external_id) ? (const xmlChar *)StringValueCStr(external_id) : NULL
+        );
+  rb_doc = noko_xml_document_wrap_with_init_args(klass, doc, argc, argv);
+  return rb_doc ;
+}
+
+/*
+ * call-seq:
+ *  read_io(io, url, encoding, options)
+ *
+ * Read the HTML document from +io+ with given +url+, +encoding+,
+ * and +options+.  See Nokogiri::HTML4.parse
+ */
+static VALUE
+rb_html_document_s_read_io(VALUE klass, VALUE rb_io, VALUE rb_url, VALUE rb_encoding, VALUE rb_options)
+{
+  VALUE rb_doc;
+  VALUE rb_error_list = rb_ary_new();
+  htmlDocPtr c_doc;
+  const char *c_url = NIL_P(rb_url) ? NULL : StringValueCStr(rb_url);
+  const char *c_encoding = NIL_P(rb_encoding) ? NULL : StringValueCStr(rb_encoding);
+  int options = NUM2INT(rb_options);
+
+  xmlSetStructuredErrorFunc((void *)rb_error_list, Nokogiri_error_array_pusher);
+
+  c_doc = htmlReadIO(noko_io_read, noko_io_close, (void *)rb_io, c_url, c_encoding, options);
+
+  xmlSetStructuredErrorFunc(NULL, NULL);
+
+  /*
+   * If EncodingFound has occurred in EncodingReader, make sure to do
+   * a cleanup and propagate the error.
+   */
+  if (rb_respond_to(rb_io, id_encoding_found)) {
+    VALUE encoding_found = rb_funcall(rb_io, id_encoding_found, 0);
+    if (!NIL_P(encoding_found)) {
+      xmlFreeDoc(c_doc);
+      rb_exc_raise(encoding_found);
+    }
+  }
+
+  if ((c_doc == NULL) || (!(options & XML_PARSE_RECOVER) && (RARRAY_LEN(rb_error_list) > 0))) {
+    VALUE rb_error ;
+
+    xmlFreeDoc(c_doc);
+
+    rb_error = rb_ary_entry(rb_error_list, 0);
+    if (rb_error == Qnil) {
+      rb_raise(rb_eRuntimeError, "Could not parse document");
+    } else {
+      VALUE exception_message = rb_funcall(rb_error, id_to_s, 0);
+      exception_message = rb_str_concat(rb_str_new2("Parser without recover option encountered error or warning: "),
+                                        exception_message);
+      rb_exc_raise(rb_class_new_instance(1, &exception_message, cNokogiriXmlSyntaxError));
+    }
+
+    return Qnil;
+  }
+
+  rb_doc = noko_xml_document_wrap(klass, c_doc);
+  rb_iv_set(rb_doc, "@errors", rb_error_list);
+  return rb_doc;
+}
+
+/*
+ * call-seq:
+ *  read_memory(string, url, encoding, options)
+ *
+ * Read the HTML document contained in +string+ with given +url+, +encoding+,
+ * and +options+.  See Nokogiri::HTML4.parse
+ */
+static VALUE
+rb_html_document_s_read_memory(VALUE klass, VALUE rb_html, VALUE rb_url, VALUE rb_encoding, VALUE rb_options)
+{
+  VALUE rb_doc;
+  VALUE rb_error_list = rb_ary_new();
+  htmlDocPtr c_doc;
+  const char *c_buffer = StringValuePtr(rb_html);
+  const char *c_url = NIL_P(rb_url) ? NULL : StringValueCStr(rb_url);
+  const char *c_encoding = NIL_P(rb_encoding) ? NULL : StringValueCStr(rb_encoding);
+  int html_len = (int)RSTRING_LEN(rb_html);
+  int options = NUM2INT(rb_options);
+
+  xmlSetStructuredErrorFunc((void *)rb_error_list, Nokogiri_error_array_pusher);
+
+  c_doc = htmlReadMemory(c_buffer, html_len, c_url, c_encoding, options);
+
+  xmlSetStructuredErrorFunc(NULL, NULL);
+
+  if ((c_doc == NULL) || (!(options & XML_PARSE_RECOVER) && (RARRAY_LEN(rb_error_list) > 0))) {
+    VALUE rb_error ;
+
+    xmlFreeDoc(c_doc);
+
+    rb_error = rb_ary_entry(rb_error_list, 0);
+    if (rb_error == Qnil) {
+      rb_raise(rb_eRuntimeError, "Could not parse document");
+    } else {
+      VALUE exception_message = rb_funcall(rb_error, id_to_s, 0);
+      exception_message = rb_str_concat(rb_str_new2("Parser without recover option encountered error or warning: "),
+                                        exception_message);
+      rb_exc_raise(rb_class_new_instance(1, &exception_message, cNokogiriXmlSyntaxError));
+    }
+
+    return Qnil;
+  }
+
+  rb_doc = noko_xml_document_wrap(klass, c_doc);
+  rb_iv_set(rb_doc, "@errors", rb_error_list);
+  return rb_doc;
+}
+
+/*
+ * call-seq:
+ *  type
+ *
+ * The type for this document
+ */
+static VALUE
+rb_html_document_type(VALUE self)
+{
+  htmlDocPtr doc;
+  Data_Get_Struct(self, xmlDoc, doc);
+  return INT2NUM((long)doc->type);
+}
+
+void
+noko_init_html_document()
+{
+  assert(cNokogiriXmlDocument);
+  cNokogiriHtml4Document = rb_define_class_under(mNokogiriHtml4, "Document", cNokogiriXmlDocument);
+
+  rb_define_singleton_method(cNokogiriHtml4Document, "read_memory", rb_html_document_s_read_memory, 4);
+  rb_define_singleton_method(cNokogiriHtml4Document, "read_io", rb_html_document_s_read_io, 4);
+  rb_define_singleton_method(cNokogiriHtml4Document, "new", rb_html_document_s_new, -1);
+
+  rb_define_method(cNokogiriHtml4Document, "type", rb_html_document_type, 0);
+
+  id_encoding_found = rb_intern("encoding_found");
+  id_to_s = rb_intern("to_s");
+}

data/ext/nokogiri/html4_element_description.c

@@ -0,0 +1,294 @@
+#include <nokogiri.h>
+
+VALUE cNokogiriHtml4ElementDescription ;
+
+/*
+ * call-seq:
+ *  required_attributes
+ *
+ * A list of required attributes for this element
+ */
+static VALUE
+required_attributes(VALUE self)
+{
+  const htmlElemDesc *description;
+  VALUE list;
+  int i;
+
+  Data_Get_Struct(self, htmlElemDesc, description);
+
+  list = rb_ary_new();
+
+  if (NULL == description->attrs_req) { return list; }
+
+  for (i = 0; description->attrs_depr[i]; i++) {
+    rb_ary_push(list, NOKOGIRI_STR_NEW2(description->attrs_req[i]));
+  }
+
+  return list;
+}
+
+/*
+ * call-seq:
+ *  deprecated_attributes
+ *
+ * A list of deprecated attributes for this element
+ */
+static VALUE
+deprecated_attributes(VALUE self)
+{
+  const htmlElemDesc *description;
+  VALUE list;
+  int i;
+
+  Data_Get_Struct(self, htmlElemDesc, description);
+
+  list = rb_ary_new();
+
+  if (NULL == description->attrs_depr) { return list; }
+
+  for (i = 0; description->attrs_depr[i]; i++) {
+    rb_ary_push(list, NOKOGIRI_STR_NEW2(description->attrs_depr[i]));
+  }
+
+  return list;
+}
+
+/*
+ * call-seq:
+ *  optional_attributes
+ *
+ * A list of optional attributes for this element
+ */
+static VALUE
+optional_attributes(VALUE self)
+{
+  const htmlElemDesc *description;
+  VALUE list;
+  int i;
+
+  Data_Get_Struct(self, htmlElemDesc, description);
+
+  list = rb_ary_new();
+
+  if (NULL == description->attrs_opt) { return list; }
+
+  for (i = 0; description->attrs_opt[i]; i++) {
+    rb_ary_push(list, NOKOGIRI_STR_NEW2(description->attrs_opt[i]));
+  }
+
+  return list;
+}
+
+/*
+ * call-seq:
+ *  default_sub_element
+ *
+ * The default sub element for this element
+ */
+static VALUE
+default_sub_element(VALUE self)
+{
+  const htmlElemDesc *description;
+  Data_Get_Struct(self, htmlElemDesc, description);
+
+  if (description->defaultsubelt) {
+    return NOKOGIRI_STR_NEW2(description->defaultsubelt);
+  }
+
+  return Qnil;
+}
+
+/*
+ * call-seq:
+ *  sub_elements
+ *
+ * A list of allowed sub elements for this element.
+ */
+static VALUE
+sub_elements(VALUE self)
+{
+  const htmlElemDesc *description;
+  VALUE list;
+  int i;
+
+  Data_Get_Struct(self, htmlElemDesc, description);
+
+  list = rb_ary_new();
+
+  if (NULL == description->subelts) { return list; }
+
+  for (i = 0; description->subelts[i]; i++) {
+    rb_ary_push(list, NOKOGIRI_STR_NEW2(description->subelts[i]));
+  }
+
+  return list;
+}
+
+/*
+ * call-seq:
+ *  description
+ *
+ * The description for this element
+ */
+static VALUE
+description(VALUE self)
+{
+  const htmlElemDesc *description;
+  Data_Get_Struct(self, htmlElemDesc, description);
+
+  return NOKOGIRI_STR_NEW2(description->desc);
+}
+
+/*
+ * call-seq:
+ *  inline?
+ *
+ * Is this element an inline element?
+ */
+static VALUE
+inline_eh(VALUE self)
+{
+  const htmlElemDesc *description;
+  Data_Get_Struct(self, htmlElemDesc, description);
+
+  if (description->isinline) { return Qtrue; }
+  return Qfalse;
+}
+
+/*
+ * call-seq:
+ *  deprecated?
+ *
+ * Is this element deprecated?
+ */
+static VALUE
+deprecated_eh(VALUE self)
+{
+  const htmlElemDesc *description;
+  Data_Get_Struct(self, htmlElemDesc, description);
+
+  if (description->depr) { return Qtrue; }
+  return Qfalse;
+}
+
+/*
+ * call-seq:
+ *  empty?
+ *
+ * Is this an empty element?
+ */
+static VALUE
+empty_eh(VALUE self)
+{
+  const htmlElemDesc *description;
+  Data_Get_Struct(self, htmlElemDesc, description);
+
+  if (description->empty) { return Qtrue; }
+  return Qfalse;
+}
+
+/*
+ * call-seq:
+ *  save_end_tag?
+ *
+ * Should the end tag be saved?
+ */
+static VALUE
+save_end_tag_eh(VALUE self)
+{
+  const htmlElemDesc *description;
+  Data_Get_Struct(self, htmlElemDesc, description);
+
+  if (description->saveEndTag) { return Qtrue; }
+  return Qfalse;
+}
+
+/*
+ * call-seq:
+ *  implied_end_tag?
+ *
+ * Can the end tag be implied for this tag?
+ */
+static VALUE
+implied_end_tag_eh(VALUE self)
+{
+  const htmlElemDesc *description;
+  Data_Get_Struct(self, htmlElemDesc, description);
+
+  if (description->endTag) { return Qtrue; }
+  return Qfalse;
+}
+
+/*
+ * call-seq:
+ *  implied_start_tag?
+ *
+ * Can the start tag be implied for this tag?
+ */
+static VALUE
+implied_start_tag_eh(VALUE self)
+{
+  const htmlElemDesc *description;
+  Data_Get_Struct(self, htmlElemDesc, description);
+
+  if (description->startTag) { return Qtrue; }
+  return Qfalse;
+}
+
+/*
+ * call-seq:
+ *  name
+ *
+ * Get the tag name for this ElemementDescription
+ */
+static VALUE
+name(VALUE self)
+{
+  const htmlElemDesc *description;
+  Data_Get_Struct(self, htmlElemDesc, description);
+
+  if (NULL == description->name) { return Qnil; }
+  return NOKOGIRI_STR_NEW2(description->name);
+}
+
+/*
+ * call-seq:
+ *  [](tag_name)
+ *
+ * Get ElemementDescription for +tag_name+
+ */
+static VALUE
+get_description(VALUE klass, VALUE tag_name)
+{
+  const htmlElemDesc *description = htmlTagLookup(
+                                      (const xmlChar *)StringValueCStr(tag_name)
+                                    );
+
+  if (NULL == description) { return Qnil; }
+  return Data_Wrap_Struct(klass, 0, 0, (void *)(uintptr_t)description);
+}
+
+void
+noko_init_html_element_description()
+{
+  cNokogiriHtml4ElementDescription = rb_define_class_under(mNokogiriHtml4, "ElementDescription", rb_cObject);
+
+  rb_undef_alloc_func(cNokogiriHtml4ElementDescription);
+
+  rb_define_singleton_method(cNokogiriHtml4ElementDescription, "[]", get_description, 1);
+
+  rb_define_method(cNokogiriHtml4ElementDescription, "name", name, 0);
+  rb_define_method(cNokogiriHtml4ElementDescription, "implied_start_tag?", implied_start_tag_eh, 0);
+  rb_define_method(cNokogiriHtml4ElementDescription, "implied_end_tag?", implied_end_tag_eh, 0);
+  rb_define_method(cNokogiriHtml4ElementDescription, "save_end_tag?", save_end_tag_eh, 0);
+  rb_define_method(cNokogiriHtml4ElementDescription, "empty?", empty_eh, 0);
+  rb_define_method(cNokogiriHtml4ElementDescription, "deprecated?", deprecated_eh, 0);
+  rb_define_method(cNokogiriHtml4ElementDescription, "inline?", inline_eh, 0);
+  rb_define_method(cNokogiriHtml4ElementDescription, "description", description, 0);
+  rb_define_method(cNokogiriHtml4ElementDescription, "sub_elements", sub_elements, 0);
+  rb_define_method(cNokogiriHtml4ElementDescription, "default_sub_element", default_sub_element, 0);
+  rb_define_method(cNokogiriHtml4ElementDescription, "optional_attributes", optional_attributes, 0);
+  rb_define_method(cNokogiriHtml4ElementDescription, "deprecated_attributes", deprecated_attributes, 0);
+  rb_define_method(cNokogiriHtml4ElementDescription, "required_attributes", required_attributes, 0);
+}