duckdb 0.8.2-dev87.0 → 0.9.1-dev0.0

package/src/duckdb/src/execution/operator/csv_scanner/buffered_csv_reader.cpp

@@ -0,0 +1,434 @@
+#include "duckdb/execution/operator/scan/csv/buffered_csv_reader.hpp"
+
+#include "duckdb/catalog/catalog_entry/table_catalog_entry.hpp"
+#include "duckdb/common/file_system.hpp"
+#include "duckdb/common/string_util.hpp"
+#include "duckdb/common/to_string.hpp"
+#include "duckdb/common/types/cast_helpers.hpp"
+#include "duckdb/common/vector_operations/unary_executor.hpp"
+#include "duckdb/common/vector_operations/vector_operations.hpp"
+#include "duckdb/execution/operator/scan/csv/csv_sniffer.hpp"
+#include "duckdb/execution/operator/scan/csv/csv_state_machine.hpp"
+#include "duckdb/function/scalar/strftime_format.hpp"
+#include "duckdb/main/client_data.hpp"
+#include "duckdb/main/database.hpp"
+#include "duckdb/main/error_manager.hpp"
+#include "duckdb/parser/column_definition.hpp"
+#include "duckdb/parser/keyword_helper.hpp"
+#include "duckdb/storage/data_table.hpp"
+#include "utf8proc.hpp"
+#include "utf8proc_wrapper.hpp"
+
+#include <algorithm>
+#include <cctype>
+#include <cstring>
+#include <fstream>
+
+namespace duckdb {
+
+BufferedCSVReader::BufferedCSVReader(ClientContext &context, CSVReaderOptions options_p,
+                                     const vector<LogicalType> &requested_types)
+    : BaseCSVReader(context, std::move(options_p), requested_types), buffer_size(0), position(0), start(0) {
+	file_handle = OpenCSV(context, options);
+	Initialize(requested_types);
+}
+
+BufferedCSVReader::BufferedCSVReader(ClientContext &context, string filename, CSVReaderOptions options_p,
+                                     const vector<LogicalType> &requested_types)
+    : BaseCSVReader(context, std::move(options_p), requested_types), buffer_size(0), position(0), start(0) {
+	options.file_path = std::move(filename);
+	file_handle = OpenCSV(context, options);
+	Initialize(requested_types);
+}
+
+void BufferedCSVReader::Initialize(const vector<LogicalType> &requested_types) {
+	if (options.auto_detect && options.file_options.union_by_name) {
+		// This is required for the sniffer to work on Union By Name
+		D_ASSERT(options.file_path == file_handle->GetFilePath());
+		auto bm_file_handle = BaseCSVReader::OpenCSV(context, options);
+		auto csv_buffer_manager = make_shared<CSVBufferManager>(context, std::move(bm_file_handle), options);
+		CSVSniffer sniffer(options, csv_buffer_manager, state_machine_cache);
+		auto sniffer_result = sniffer.SniffCSV();
+		return_types = sniffer_result.return_types;
+		names = sniffer_result.names;
+		if (return_types.empty()) {
+			throw InvalidInputException("Failed to detect column types from CSV: is the file a valid CSV file?");
+		}
+	} else {
+		return_types = requested_types;
+		ResetBuffer();
+	}
+	SkipRowsAndReadHeader(options.dialect_options.skip_rows, options.dialect_options.header);
+	InitParseChunk(return_types.size());
+}
+
+void BufferedCSVReader::ResetBuffer() {
+	buffer.reset();
+	buffer_size = 0;
+	position = 0;
+	start = 0;
+	cached_buffers.clear();
+}
+
+void BufferedCSVReader::SkipRowsAndReadHeader(idx_t skip_rows, bool skip_header) {
+	for (idx_t i = 0; i < skip_rows; i++) {
+		// ignore skip rows
+		string read_line = file_handle->ReadLine();
+		linenr++;
+	}
+
+	if (skip_header) {
+		// ignore the first line as a header line
+		InitParseChunk(return_types.size());
+		ParseCSV(ParserMode::PARSING_HEADER);
+	}
+}
+
+string BufferedCSVReader::ColumnTypesError(case_insensitive_map_t<idx_t> sql_types_per_column,
+                                           const vector<string> &names) {
+	for (idx_t i = 0; i < names.size(); i++) {
+		auto it = sql_types_per_column.find(names[i]);
+		if (it != sql_types_per_column.end()) {
+			sql_types_per_column.erase(names[i]);
+			continue;
+		}
+	}
+	if (sql_types_per_column.empty()) {
+		return string();
+	}
+	string exception = "COLUMN_TYPES error: Columns with names: ";
+	for (auto &col : sql_types_per_column) {
+		exception += "\"" + col.first + "\",";
+	}
+	exception.pop_back();
+	exception += " do not exist in the CSV File";
+	return exception;
+}
+
+void BufferedCSVReader::SkipEmptyLines() {
+	if (parse_chunk.data.size() == 1) {
+		// Empty lines are null data.
+		return;
+	}
+	for (; position < buffer_size; position++) {
+		if (!StringUtil::CharacterIsNewline(buffer[position])) {
+			return;
+		}
+	}
+}
+
+void UpdateMaxLineLength(ClientContext &context, idx_t line_length) {
+	if (!context.client_data->debug_set_max_line_length) {
+		return;
+	}
+	if (line_length < context.client_data->debug_max_line_length) {
+		return;
+	}
+	context.client_data->debug_max_line_length = line_length;
+}
+
+bool BufferedCSVReader::ReadBuffer(idx_t &start, idx_t &line_start) {
+	if (start > buffer_size) {
+		return false;
+	}
+	auto old_buffer = std::move(buffer);
+
+	// the remaining part of the last buffer
+	idx_t remaining = buffer_size - start;
+
+	idx_t buffer_read_size = INITIAL_BUFFER_SIZE_LARGE;
+
+	while (remaining > buffer_read_size) {
+		buffer_read_size *= 2;
+	}
+
+	// Check line length
+	if (remaining > options.maximum_line_size) {
+		throw InvalidInputException("Maximum line size of %llu bytes exceeded on line %s!", options.maximum_line_size,
+		                            GetLineNumberStr(linenr, linenr_estimated));
+	}
+
+	buffer = make_unsafe_uniq_array<char>(buffer_read_size + remaining + 1);
+	buffer_size = remaining + buffer_read_size;
+	if (remaining > 0) {
+		// remaining from last buffer: copy it here
+		memcpy(buffer.get(), old_buffer.get() + start, remaining);
+	}
+	idx_t read_count = file_handle->Read(buffer.get() + remaining, buffer_read_size);
+
+	bytes_in_chunk += read_count;
+	buffer_size = remaining + read_count;
+	buffer[buffer_size] = '\0';
+	if (old_buffer) {
+		cached_buffers.push_back(std::move(old_buffer));
+	}
+	start = 0;
+	position = remaining;
+	if (!bom_checked) {
+		bom_checked = true;
+		if (read_count >= 3 && buffer[0] == '\xEF' && buffer[1] == '\xBB' && buffer[2] == '\xBF') {
+			start += 3;
+			position += 3;
+		}
+	}
+	line_start = start;
+
+	return read_count > 0;
+}
+
+void BufferedCSVReader::ParseCSV(DataChunk &insert_chunk) {
+	string error_message;
+	if (!TryParseCSV(ParserMode::PARSING, insert_chunk, error_message)) {
+		throw InvalidInputException(error_message);
+	}
+}
+
+void BufferedCSVReader::ParseCSV(ParserMode mode) {
+	DataChunk dummy_chunk;
+	string error_message;
+	if (!TryParseCSV(mode, dummy_chunk, error_message)) {
+		throw InvalidInputException(error_message);
+	}
+}
+
+bool BufferedCSVReader::TryParseCSV(ParserMode parser_mode, DataChunk &insert_chunk, string &error_message) {
+	mode = parser_mode;
+	// used for parsing algorithm
+	bool finished_chunk = false;
+	idx_t column = 0;
+	idx_t offset = 0;
+	bool has_quotes = false;
+	vector<idx_t> escape_positions;
+
+	idx_t line_start = position;
+	idx_t line_size = 0;
+	// read values into the buffer (if any)
+	if (position >= buffer_size) {
+		if (!ReadBuffer(start, line_start)) {
+			return true;
+		}
+	}
+
+	// start parsing the first value
+	goto value_start;
+value_start:
+	offset = 0;
+	/* state: value_start */
+	// this state parses the first character of a value
+	if (buffer[position] == options.dialect_options.state_machine_options.quote) {
+		// quote: actual value starts in the next position
+		// move to in_quotes state
+		start = position + 1;
+		line_size++;
+		goto in_quotes;
+	} else {
+		// no quote, move to normal parsing state
+		start = position;
+		goto normal;
+	}
+normal:
+	/* state: normal parsing state */
+	// this state parses the remainder of a non-quoted value until we reach a delimiter or newline
+	do {
+		for (; position < buffer_size; position++) {
+			line_size++;
+			if (buffer[position] == options.dialect_options.state_machine_options.delimiter) {
+				// delimiter: end the value and add it to the chunk
+				goto add_value;
+			} else if (StringUtil::CharacterIsNewline(buffer[position])) {
+				// newline: add row
+				goto add_row;
+			}
+		}
+	} while (ReadBuffer(start, line_start));
+	// file ends during normal scan: go to end state
+	goto final_state;
+add_value:
+	AddValue(string_t(buffer.get() + start, position - start - offset), column, escape_positions, has_quotes);
+	// increase position by 1 and move start to the new position
+	offset = 0;
+	has_quotes = false;
+	start = ++position;
+	line_size++;
+	if (position >= buffer_size && !ReadBuffer(start, line_start)) {
+		// file ends right after delimiter, go to final state
+		goto final_state;
+	}
+	goto value_start;
+add_row : {
+	// check type of newline (\r or \n)
+	bool carriage_return = buffer[position] == '\r';
+	AddValue(string_t(buffer.get() + start, position - start - offset), column, escape_positions, has_quotes);
+	if (!error_message.empty()) {
+		return false;
+	}
+	VerifyLineLength(position - line_start);
+
+	finished_chunk = AddRow(insert_chunk, column, error_message);
+	UpdateMaxLineLength(context, position - line_start);
+	if (!error_message.empty()) {
+		return false;
+	}
+	// increase position by 1 and move start to the new position
+	offset = 0;
+	has_quotes = false;
+	position++;
+	line_size = 0;
+	start = position;
+	line_start = position;
+	if (position >= buffer_size && !ReadBuffer(start, line_start)) {
+		// file ends right after delimiter, go to final state
+		goto final_state;
+	}
+	if (carriage_return) {
+		// \r newline, go to special state that parses an optional \n afterwards
+		goto carriage_return;
+	} else {
+		SetNewLineDelimiter();
+		SkipEmptyLines();
+
+		start = position;
+		line_start = position;
+		if (position >= buffer_size && !ReadBuffer(start, line_start)) {
+			// file ends right after delimiter, go to final state
+			goto final_state;
+		}
+		// \n newline, move to value start
+		if (finished_chunk) {
+			return true;
+		}
+		goto value_start;
+	}
+}
+in_quotes:
+	/* state: in_quotes */
+	// this state parses the remainder of a quoted value
+	has_quotes = true;
+	position++;
+	line_size++;
+	do {
+		for (; position < buffer_size; position++) {
+			line_size++;
+			if (buffer[position] == options.dialect_options.state_machine_options.quote) {
+				// quote: move to unquoted state
+				goto unquote;
+			} else if (buffer[position] == options.dialect_options.state_machine_options.escape) {
+				// escape: store the escaped position and move to handle_escape state
+				escape_positions.push_back(position - start);
+				goto handle_escape;
+			}
+		}
+	} while (ReadBuffer(start, line_start));
+	// still in quoted state at the end of the file, error:
+	throw InvalidInputException("Error in file \"%s\" on line %s: unterminated quotes. (%s)", options.file_path,
+	                            GetLineNumberStr(linenr, linenr_estimated).c_str(), options.ToString());
+unquote:
+	/* state: unquote */
+	// this state handles the state directly after we unquote
+	// in this state we expect either another quote (entering the quoted state again, and escaping the quote)
+	// or a delimiter/newline, ending the current value and moving on to the next value
+	position++;
+	line_size++;
+	if (position >= buffer_size && !ReadBuffer(start, line_start)) {
+		// file ends right after unquote, go to final state
+		offset = 1;
+		goto final_state;
+	}
+	if (buffer[position] == options.dialect_options.state_machine_options.quote &&
+	    (options.dialect_options.state_machine_options.escape == '\0' ||
+	     options.dialect_options.state_machine_options.escape == options.dialect_options.state_machine_options.quote)) {
+		// escaped quote, return to quoted state and store escape position
+		escape_positions.push_back(position - start);
+		goto in_quotes;
+	} else if (buffer[position] == options.dialect_options.state_machine_options.delimiter) {
+		// delimiter, add value
+		offset = 1;
+		goto add_value;
+	} else if (StringUtil::CharacterIsNewline(buffer[position])) {
+		offset = 1;
+		goto add_row;
+	} else {
+		error_message = StringUtil::Format(
+		    "Error in file \"%s\" on line %s: quote should be followed by end of value, end of "
+		    "row or another quote. (%s)",
+		    options.file_path, GetLineNumberStr(linenr, linenr_estimated).c_str(), options.ToString());
+		return false;
+	}
+handle_escape:
+	/* state: handle_escape */
+	// escape should be followed by a quote or another escape character
+	position++;
+	line_size++;
+	if (position >= buffer_size && !ReadBuffer(start, line_start)) {
+		error_message = StringUtil::Format(
+		    "Error in file \"%s\" on line %s: neither QUOTE nor ESCAPE is proceeded by ESCAPE. (%s)", options.file_path,
+		    GetLineNumberStr(linenr, linenr_estimated).c_str(), options.ToString());
+		return false;
+	}
+	if (buffer[position] != options.dialect_options.state_machine_options.quote &&
+	    buffer[position] != options.dialect_options.state_machine_options.escape) {
+		error_message = StringUtil::Format(
+		    "Error in file \"%s\" on line %s: neither QUOTE nor ESCAPE is proceeded by ESCAPE. (%s)", options.file_path,
+		    GetLineNumberStr(linenr, linenr_estimated).c_str(), options.ToString());
+		return false;
+	}
+	// escape was followed by quote or escape, go back to quoted state
+	goto in_quotes;
+carriage_return:
+	/* state: carriage_return */
+	// this stage optionally skips a newline (\n) character, which allows \r\n to be interpreted as a single line
+	if (buffer[position] == '\n') {
+		SetNewLineDelimiter(true, true);
+		// newline after carriage return: skip
+		// increase position by 1 and move start to the new position
+		start = ++position;
+		line_size++;
+
+		if (position >= buffer_size && !ReadBuffer(start, line_start)) {
+			// file ends right after delimiter, go to final state
+			goto final_state;
+		}
+	} else {
+		SetNewLineDelimiter(true, false);
+	}
+	if (finished_chunk) {
+		return true;
+	}
+	SkipEmptyLines();
+	start = position;
+	line_start = position;
+	if (position >= buffer_size && !ReadBuffer(start, line_start)) {
+		// file ends right after delimiter, go to final state
+		goto final_state;
+	}
+
+	goto value_start;
+final_state:
+	if (finished_chunk) {
+		return true;
+	}
+
+	if (column > 0 || position > start) {
+		// remaining values to be added to the chunk
+		AddValue(string_t(buffer.get() + start, position - start - offset), column, escape_positions, has_quotes);
+		VerifyLineLength(position - line_start);
+
+		finished_chunk = AddRow(insert_chunk, column, error_message);
+		SkipEmptyLines();
+		UpdateMaxLineLength(context, line_size);
+		if (!error_message.empty()) {
+			return false;
+		}
+	}
+
+	// final stage, only reached after parsing the file is finished
+	// flush the parsed chunk and finalize parsing
+	if (mode == ParserMode::PARSING) {
+		Flush(insert_chunk);
+	}
+
+	end_of_file_reached = true;
+	return true;
+}
+
+} // namespace duckdb

package/src/duckdb/src/execution/operator/csv_scanner/csv_buffer.cpp

@@ -0,0 +1,89 @@
+#include "duckdb/execution/operator/scan/csv/csv_buffer.hpp"
+#include "duckdb/common/string_util.hpp"
+
+namespace duckdb {
+
+CSVBuffer::CSVBuffer(ClientContext &context, idx_t buffer_size_p, CSVFileHandle &file_handle,
+                     idx_t &global_csv_current_position, idx_t file_number_p)
+    : context(context), first_buffer(true), file_number(file_number_p), can_seek(file_handle.CanSeek()) {
+	AllocateBuffer(buffer_size_p);
+	auto buffer = Ptr();
+	actual_buffer_size = file_handle.Read(buffer, buffer_size_p);
+	while (actual_buffer_size < buffer_size_p && !file_handle.FinishedReading()) {
+		// We keep reading until this block is full
+		actual_buffer_size += file_handle.Read(&buffer[actual_buffer_size], buffer_size_p - actual_buffer_size);
+	}
+	global_csv_start = global_csv_current_position;
+	// BOM check (https://en.wikipedia.org/wiki/Byte_order_mark)
+	if (actual_buffer_size >= 3 && buffer[0] == '\xEF' && buffer[1] == '\xBB' && buffer[2] == '\xBF') {
+		start_position += 3;
+	}
+	last_buffer = file_handle.FinishedReading();
+}
+
+CSVBuffer::CSVBuffer(CSVFileHandle &file_handle, ClientContext &context, idx_t buffer_size,
+                     idx_t global_csv_current_position, idx_t file_number_p)
+    : context(context), global_csv_start(global_csv_current_position), file_number(file_number_p),
+      can_seek(file_handle.CanSeek()) {
+	AllocateBuffer(buffer_size);
+	auto buffer = handle.Ptr();
+	actual_buffer_size = file_handle.Read(handle.Ptr(), buffer_size);
+	while (actual_buffer_size < buffer_size && !file_handle.FinishedReading()) {
+		// We keep reading until this block is full
+		actual_buffer_size += file_handle.Read(&buffer[actual_buffer_size], buffer_size - actual_buffer_size);
+	}
+	last_buffer = file_handle.FinishedReading();
+}
+
+shared_ptr<CSVBuffer> CSVBuffer::Next(CSVFileHandle &file_handle, idx_t buffer_size, idx_t file_number_p) {
+	auto next_csv_buffer =
+	    make_shared<CSVBuffer>(file_handle, context, buffer_size, global_csv_start + actual_buffer_size, file_number_p);
+	if (next_csv_buffer->GetBufferSize() == 0) {
+		// We are done reading
+		return nullptr;
+	}
+	return next_csv_buffer;
+}
+
+void CSVBuffer::AllocateBuffer(idx_t buffer_size) {
+	auto &buffer_manager = BufferManager::GetBufferManager(context);
+	bool can_destroy = can_seek;
+	handle = buffer_manager.Allocate(MaxValue<idx_t>(Storage::BLOCK_SIZE, buffer_size), can_destroy, &block);
+}
+
+idx_t CSVBuffer::GetBufferSize() {
+	return actual_buffer_size;
+}
+
+void CSVBuffer::Reload(CSVFileHandle &file_handle) {
+	AllocateBuffer(actual_buffer_size);
+	file_handle.Seek(global_csv_start);
+	file_handle.Read(handle.Ptr(), actual_buffer_size);
+}
+
+unique_ptr<CSVBufferHandle> CSVBuffer::Pin(CSVFileHandle &file_handle) {
+	auto &buffer_manager = BufferManager::GetBufferManager(context);
+	if (can_seek && block->IsUnloaded()) {
+		// We have to reload it from disk
+		block = nullptr;
+		Reload(file_handle);
+	}
+	return make_uniq<CSVBufferHandle>(buffer_manager.Pin(block), actual_buffer_size, first_buffer, last_buffer,
+	                                  global_csv_start, start_position, file_number);
+}
+
+void CSVBuffer::Unpin() {
+	if (handle.IsValid()) {
+		handle.Destroy();
+	}
+}
+
+idx_t CSVBuffer::GetStart() {
+	return start_position;
+}
+
+bool CSVBuffer::IsCSVFileLastBuffer() {
+	return last_buffer;
+}
+
+} // namespace duckdb

package/src/duckdb/src/execution/operator/csv_scanner/csv_buffer_manager.cpp

@@ -0,0 +1,90 @@
+#include "duckdb/execution/operator/scan/csv/csv_buffer_manager.hpp"
+#include "duckdb/execution/operator/scan/csv/csv_buffer.hpp"
+namespace duckdb {
+
+CSVBufferManager::CSVBufferManager(ClientContext &context_p, unique_ptr<CSVFileHandle> file_handle_p,
+                                   const CSVReaderOptions &options, idx_t file_idx_p)
+    : file_handle(std::move(file_handle_p)), context(context_p), file_idx(file_idx_p),
+      buffer_size(CSVBuffer::CSV_BUFFER_SIZE) {
+	if (options.skip_rows_set) {
+		// Skip rows if they are set
+		skip_rows = options.dialect_options.skip_rows;
+	}
+	auto file_size = file_handle->FileSize();
+	if (file_size > 0 && file_size < buffer_size) {
+		buffer_size = CSVBuffer::CSV_MINIMUM_BUFFER_SIZE;
+	}
+	if (options.buffer_size < buffer_size) {
+		buffer_size = options.buffer_size;
+	}
+	for (idx_t i = 0; i < skip_rows; i++) {
+		file_handle->ReadLine();
+	}
+	Initialize();
+}
+
+void CSVBufferManager::UnpinBuffer(idx_t cache_idx) {
+	if (cache_idx < cached_buffers.size()) {
+		cached_buffers[cache_idx]->Unpin();
+	}
+}
+
+void CSVBufferManager::Initialize() {
+	if (cached_buffers.empty()) {
+		cached_buffers.emplace_back(
+		    make_shared<CSVBuffer>(context, buffer_size, *file_handle, global_csv_pos, file_idx));
+		last_buffer = cached_buffers.front();
+	}
+	start_pos = last_buffer->GetStart();
+}
+
+idx_t CSVBufferManager::GetStartPos() {
+	return start_pos;
+}
+bool CSVBufferManager::ReadNextAndCacheIt() {
+	D_ASSERT(last_buffer);
+	if (!last_buffer->IsCSVFileLastBuffer()) {
+		auto maybe_last_buffer = last_buffer->Next(*file_handle, buffer_size, file_idx);
+		if (!maybe_last_buffer) {
+			last_buffer->last_buffer = true;
+			return false;
+		}
+		last_buffer = std::move(maybe_last_buffer);
+		cached_buffers.emplace_back(last_buffer);
+		return true;
+	}
+	return false;
+}
+
+unique_ptr<CSVBufferHandle> CSVBufferManager::GetBuffer(const idx_t pos) {
+	while (pos >= cached_buffers.size()) {
+		if (done) {
+			return nullptr;
+		}
+		if (!ReadNextAndCacheIt()) {
+			done = true;
+		}
+	}
+	if (pos != 0) {
+		cached_buffers[pos - 1]->Unpin();
+	}
+	return cached_buffers[pos]->Pin(*file_handle);
+}
+
+bool CSVBufferIterator::Finished() {
+	return !cur_buffer_handle;
+}
+
+void CSVBufferIterator::Reset() {
+	if (cur_buffer_handle) {
+		cur_buffer_handle.reset();
+	}
+	if (cur_buffer_idx > 0) {
+		buffer_manager->UnpinBuffer(cur_buffer_idx - 1);
+	}
+	cur_buffer_idx = 0;
+	buffer_manager->Initialize();
+	cur_pos = buffer_manager->GetStartPos();
+}
+
+} // namespace duckdb

package/src/duckdb/src/execution/operator/csv_scanner/csv_file_handle.cpp

@@ -0,0 +1,95 @@
+#include "duckdb/execution/operator/scan/csv/csv_file_handle.hpp"
+
+namespace duckdb {
+
+CSVFileHandle::CSVFileHandle(FileSystem &fs, Allocator &allocator, unique_ptr<FileHandle> file_handle_p,
+                             const string &path_p, FileCompressionType compression)
+    : file_handle(std::move(file_handle_p)), path(path_p) {
+	can_seek = file_handle->CanSeek();
+	on_disk_file = file_handle->OnDiskFile();
+	file_size = file_handle->GetFileSize();
+}
+
+unique_ptr<FileHandle> CSVFileHandle::OpenFileHandle(FileSystem &fs, Allocator &allocator, const string &path,
+                                                     FileCompressionType compression) {
+	auto file_handle = fs.OpenFile(path, FileFlags::FILE_FLAGS_READ, FileLockType::NO_LOCK, compression);
+	if (file_handle->CanSeek()) {
+		file_handle->Reset();
+	}
+	return file_handle;
+}
+
+unique_ptr<CSVFileHandle> CSVFileHandle::OpenFile(FileSystem &fs, Allocator &allocator, const string &path,
+                                                  FileCompressionType compression) {
+	auto file_handle = CSVFileHandle::OpenFileHandle(fs, allocator, path, compression);
+	return make_uniq<CSVFileHandle>(fs, allocator, std::move(file_handle), path, compression);
+}
+
+bool CSVFileHandle::CanSeek() {
+	return can_seek;
+}
+
+void CSVFileHandle::Seek(idx_t position) {
+	if (!can_seek) {
+		throw InternalException("Cannot seek in this file");
+	}
+	file_handle->Seek(position);
+}
+
+bool CSVFileHandle::OnDiskFile() {
+	return on_disk_file;
+}
+
+idx_t CSVFileHandle::FileSize() {
+	return file_size;
+}
+
+bool CSVFileHandle::FinishedReading() {
+	return finished;
+}
+
+idx_t CSVFileHandle::Read(void *buffer, idx_t nr_bytes) {
+	requested_bytes += nr_bytes;
+	// if this is a plain file source OR we can seek we are not caching anything
+	auto bytes_read = file_handle->Read(buffer, nr_bytes);
+	if (!finished) {
+		finished = bytes_read == 0;
+	}
+	return bytes_read;
+}
+
+string CSVFileHandle::ReadLine() {
+	bool carriage_return = false;
+	string result;
+	char buffer[1];
+	while (true) {
+		idx_t bytes_read = Read(buffer, 1);
+		if (bytes_read == 0) {
+			return result;
+		}
+		if (carriage_return) {
+			if (buffer[0] != '\n') {
+				if (!file_handle->CanSeek()) {
+					throw BinderException(
+					    "Carriage return newlines not supported when reading CSV files in which we cannot seek");
+				}
+				file_handle->Seek(file_handle->SeekPosition() - 1);
+				return result;
+			}
+		}
+		if (buffer[0] == '\n') {
+			return result;
+		}
+		if (buffer[0] != '\r') {
+			result += buffer[0];
+		} else {
+			carriage_return = true;
+		}
+	}
+}
+
+string CSVFileHandle::GetFilePath() {
+	return path;
+}
+
+} // namespace duckdb