Import beginnings of a CSS parsing library.

Currently comprises a lexer.

svn path=/trunk/libcss/; revision=4112

12 files changed, 2745 insertions, 0 deletions

diff --git a/src/Makefile b/src/Makefile
new file mode 100644
index 0000000..f56a87f
--- /dev/null
+++ b/src/Makefile
@@ -0,0 +1,49 @@
+# Child makefile fragment
+#
+# Toolchain is provided by top-level makefile
+#
+# Variables provided by top-level makefile
+#
+# COMPONENT		The name of the component
+# EXPORT		The location of the export directory
+# TOP			The location of the source tree root
+# RELEASEDIR		The place to put release objects
+# DEBUGDIR		The place to put debug objects
+#
+# do_include		Canned command sequence to include a child makefile
+#
+# Variables provided by parent makefile:
+#
+# DIR			The name of the directory we're in, relative to $(TOP)
+#
+# Variables we can manipulate:
+#
+# ITEMS_CLEAN		The list of items to remove for "make clean"
+# ITEMS_DISTCLEAN	The list of items to remove for "make distclean"
+# TARGET_TESTS		The list of target names to run for "make test"
+#
+# SOURCES		The list of sources to build for $(COMPONENT)
+#
+# Plus anything from the toolchain
+
+# Push parent directory onto the directory stack
+sp             := $(sp).x
+dirstack_$(sp) := $(d)
+d              := $(DIR)
+
+# Manipulate include paths
+CFLAGS := $(CFLAGS) -I$(d)
+
+# Sources
+SRCS_$(d) := libcss.c
+
+# Append to sources for component
+SOURCES += $(addprefix $(d), $(SRCS_$(d)))
+
+# Now include any children we may have
+MAKE_INCLUDES := $(wildcard $(d)*/Makefile)
+$(eval $(foreach INC, $(MAKE_INCLUDES), $(call do_include,$(INC))))
+
+# Finally, pop off the directory stack
+d  := $(dirstack_$(sp))
+sp := $(basename $(sp))
diff --git a/src/charset/Makefile b/src/charset/Makefile
new file mode 100644
index 0000000..dda58c1
--- /dev/null
+++ b/src/charset/Makefile
@@ -0,0 +1,46 @@
+# Child makefile fragment
+#
+# Toolchain is provided by top-level makefile
+#
+# Variables provided by top-level makefile
+#
+# COMPONENT		The name of the component
+# EXPORT		The location of the export directory
+# TOP			The location of the source tree root
+# RELEASEDIR		The place to put release objects
+# DEBUGDIR		The place to put debug objects
+#
+# do_include		Canned command sequence to include a child makefile
+#
+# Variables provided by parent makefile:
+#
+# DIR			The name of the directory we're in, relative to $(TOP)
+#
+# Variables we can manipulate:
+#
+# ITEMS_CLEAN		The list of items to remove for "make clean"
+# ITEMS_DISTCLEAN	The list of items to remove for "make distclean"
+# TARGET_TESTS		The list of target names to run for "make test"
+#
+# SOURCES		The list of sources to build for $(COMPONENT)
+#
+# Plus anything from the toolchain
+
+# Push parent directory onto the directory stack
+sp             := $(sp).x
+dirstack_$(sp) := $(d)
+d              := $(DIR)
+
+# Sources
+SRCS_$(d) := detect.c
+
+# Append to sources for component
+SOURCES += $(addprefix $(d), $(SRCS_$(d)))
+
+# Now include any children we may have
+MAKE_INCLUDES := $(wildcard $(d)*/Makefile)
+$(eval $(foreach INC, $(MAKE_INCLUDES), $(call do_include,$(INC))))
+
+# Finally, pop off the directory stack
+d  := $(dirstack_$(sp))
+sp := $(basename $(sp))
diff --git a/src/charset/detect.c b/src/charset/detect.c
new file mode 100644
index 0000000..11ee699
--- /dev/null
+++ b/src/charset/detect.c
@@ -0,0 +1,161 @@
+/*
+ * This file is part of LibCSS.
+ * Licensed under the MIT License,
+ *                http://www.opensource.org/licenses/mit-license.php
+ * Copyright 2008 John-Mark Bell <jmb@netsurf-browser.org>
+ */
+
+#include <stdbool.h>
+#include <string.h>
+
+#include <parserutils/charset/mibenum.h>
+
+#include "charset/detect.h"
+#include "utils/utils.h"
+
+static parserutils_error css_charset_read_bom_or_charset(const uint8_t *data, 
+		size_t len, uint16_t *mibenum);
+
+/**
+ * Extract a charset from a chunk of data
+ *
+ * \param data     Pointer to buffer containing data
+ * \param len      Buffer length
+ * \param mibenum  Pointer to location containing current MIB enum
+ * \param source   Pointer to location containing current charset source
+ * \return PARSERUTILS_OK on success, appropriate error otherwise
+ *
+ * ::mibenum and ::source will be updated on exit
+ *
+ * CSS 2.1 $4.4
+ */
+parserutils_error css_charset_extract(const uint8_t *data, size_t len,
+		uint16_t *mibenum, uint32_t *source)
+{
+	css_error error;
+	uint16_t charset = 0;
+
+	if (data == NULL || mibenum == NULL || source == NULL)
+		return PARSERUTILS_BADPARM;
+
+	/* If the charset was dictated by the client, we've nothing to detect */
+	if (*source == CSS_CHARSET_DICTATED)
+		return PARSERUTILS_OK;
+
+	/* We need at least 4 bytes of data */
+	if (len < 4)
+		goto default_encoding;
+
+	/* Look for a BOM and/or @charset */
+	error = css_charset_read_bom_or_charset(data, len, &charset);
+	if (error != PARSERUTILS_OK)
+		return error;
+
+	if (charset != 0) {
+		*mibenum = charset;
+		*source = CSS_CHARSET_DOCUMENT;
+
+		return PARSERUTILS_OK;
+	}
+
+	/* If we've already got a charset from the linking mechanism or 
+	 * referring document, then we've nothing further to do */
+	if (*source != CSS_CHARSET_DEFAULT)
+		return PARSERUTILS_OK;
+
+	/* We've not yet found a charset, so use the default fallback */
+default_encoding:
+
+	charset = parserutils_charset_mibenum_from_name("UTF-8", SLEN("UTF-8"));
+
+	*mibenum = charset;
+	*source = CSS_CHARSET_DEFAULT;
+
+	return PARSERUTILS_OK;
+}
+
+
+/**
+ * Inspect the beginning of a buffer of data for the presence of a
+ * UTF Byte Order Mark and/or an @charset rule
+ *
+ * \param data     Pointer to buffer containing data
+ * \param len      Buffer length
+ * \param mibenum  Pointer to location to receive MIB enum
+ * \return PARSERUTILS_OK on success, appropriate error otherwise
+ */
+parserutils_error css_charset_read_bom_or_charset(const uint8_t *data, 
+		size_t len, uint16_t *mibenum)
+{
+	uint16_t charset = 0;
+
+	if (data == NULL)
+		return PARSERUTILS_BADPARM;
+
+	/* We require at least 4 bytes of data */
+	if (len < 4)
+		return PARSERUTILS_NEEDDATA;
+
+
+	/* Look for BOM */
+	if (data[0] == 0x00 && data[1] == 0x00 && 
+			data[2] == 0xFE && data[3] == 0xFF) {
+		charset = parserutils_charset_mibenum_from_name("UTF-32BE", 
+				SLEN("UTF-32BE"));
+	} else if (data[0] == 0xFF && data[1] == 0xFE &&
+			data[2] == 0x00 && data[3] == 0x00) {
+		charset = parserutils_charset_mibenum_from_name("UTF-32LE", 
+				SLEN("UTF-32LE"));
+	} else if (data[0] == 0xFE && data[1] == 0xFF) {
+		charset = parserutils_charset_mibenum_from_name("UTF-16BE", 
+				SLEN("UTF-16BE"));
+	} else if (data[0] == 0xFF && data[1] == 0xFE) {
+		charset = parserutils_charset_mibenum_from_name("UTF-16LE", 
+				SLEN("UTF-16LE"));
+	} else if (data[0] == 0xEF && data[1] == 0xBB && data[2] == 0xBF) {
+		charset = parserutils_charset_mibenum_from_name("UTF-8", 
+				SLEN("UTF-8"));
+	}
+
+	/* BOM beats @charset.
+	 * UAs differ here, but none appear to match the spec. 
+	 * The spec indicates that any @charset present in conjunction with a 
+	 * BOM, should match the BOM. In reality, it appears UAs just take the 
+	 * BOM as gospel and ignore any @charset rule. The w3c CSS validator 
+	 * appears to do the same (at the least, it doesn't complain about a
+	 * mismatch).
+	 */
+	if (charset != 0) {
+		*mibenum = charset;
+		return PARSERUTILS_OK;
+	}
+
+	/** \todo UTF-32 and UTF-16 @charset support */
+
+	/* Look for @charset, assuming ASCII-compatible source data */
+	if (len > 10 &&	strncmp((const char *) data, "@charset \"", 
+			SLEN("@charset \"")) == 0) {
+		const uint8_t *end;
+
+		/* Look for "; at end of charset declaration */
+		for (end = data + 10; end < data + len; end++) {
+			if (*end == '"' && end < data + len - 1 && 
+					*(end + 1) == ';')
+				break;
+		}
+
+		if (end == data + len) {
+			/* Ran out of input */
+			return PARSERUTILS_NEEDDATA;
+		}
+
+		/* Convert to MIB enum */
+		charset = parserutils_charset_mibenum_from_name(
+				(char *) data + 10,  end - data - 10);
+	}
+
+	*mibenum = charset;
+
+	return PARSERUTILS_OK;
+}
+
diff --git a/src/charset/detect.h b/src/charset/detect.h
new file mode 100644
index 0000000..d907921
--- /dev/null
+++ b/src/charset/detect.h
@@ -0,0 +1,24 @@
+/*
+ * This file is part of LibCSS.
+ * Licensed under the MIT License,
+ *                http://www.opensource.org/licenses/mit-license.php
+ * Copyright 2007 John-Mark Bell <jmb@netsurf-browser.org>
+ */
+
+#ifndef css_charset_detect_h_
+#define css_charset_detect_h_
+
+#include <inttypes.h>
+
+#include <libcss/errors.h>
+#include <libcss/functypes.h>
+#include <libcss/types.h>
+
+#include <parserutils/errors.h>
+
+/* Extract a charset from a chunk of data */
+parserutils_error css_charset_extract(const uint8_t *data, size_t len,
+		uint16_t *mibenum, uint32_t *source);
+
+#endif
+
diff --git a/src/lex/Makefile b/src/lex/Makefile
new file mode 100644
index 0000000..73f8ecf
--- /dev/null
+++ b/src/lex/Makefile
@@ -0,0 +1,49 @@
+# Child makefile fragment
+#
+# Toolchain is provided by top-level makefile
+#
+# Variables provided by top-level makefile
+#
+# COMPONENT		The name of the component
+# EXPORT		The location of the export directory
+# TOP			The location of the source tree root
+# RELEASEDIR		The place to put release objects
+# DEBUGDIR		The place to put debug objects
+#
+# do_include		Canned command sequence to include a child makefile
+#
+# Variables provided by parent makefile:
+#
+# DIR			The name of the directory we're in, relative to $(TOP)
+#
+# Variables we can manipulate:
+#
+# ITEMS_CLEAN		The list of items to remove for "make clean"
+# ITEMS_DISTCLEAN	The list of items to remove for "make distclean"
+# TARGET_TESTS		The list of target names to run for "make test"
+#
+# SOURCES		The list of sources to build for $(COMPONENT)
+#
+# Plus anything from the toolchain
+
+# Push parent directory onto the directory stack
+sp             := $(sp).x
+dirstack_$(sp) := $(d)
+d              := $(DIR)
+
+# Manipulate include paths
+CFLAGS := $(CFLAGS) -I$(d)
+
+# Sources
+SRCS_$(d) := lex.c
+
+# Append to sources for component
+SOURCES += $(addprefix $(d), $(SRCS_$(d)))
+
+# Now include any children we may have
+MAKE_INCLUDES := $(wildcard $(d)*/Makefile)
+$(eval $(foreach INC, $(MAKE_INCLUDES), $(call do_include,$(INC))))
+
+# Finally, pop off the directory stack
+d  := $(dirstack_$(sp))
+sp := $(basename $(sp))
diff --git a/src/lex/lex.c b/src/lex/lex.c
new file mode 100644
index 0000000..4df6cea
--- /dev/null
+++ b/src/lex/lex.c
@@ -0,0 +1,2116 @@
+/*
+ * This file is part of LibCSS.
+ * Licensed under the MIT License,
+ *                http://www.opensource.org/licenses/mit-license.php
+ * Copyright 2008 John-Mark Bell <jmb@netsurf-browser.org>
+ */
+
+/** \file CSS lexer
+ * 
+ * See docs/Tokens for the production rules used by this lexer.
+ *
+ * See docs/Lexer for the inferred first characters for each token.
+ *
+ * See also CSS3 Syntax module and CSS2.1 $4.1.1 + errata
+ *
+ * The lexer assumes that all invalid Unicode codepoints have been converted
+ * to U+FFFD by the input stream.
+ *
+ * The lexer comprises a state machine, the top-level of which is derived from
+ * the First sets in docs/Lexer. Each top-level state may contain a number of
+ * sub states. These enable restarting of the parser.
+ */
+
+#include <assert.h>
+#include <stdarg.h>
+#include <stdbool.h>
+#include <stdint.h>
+
+#include <parserutils/charset/utf8.h>
+#include <parserutils/input/inputstream.h>
+#include <parserutils/utils/buffer.h>
+
+#include <libcss/errors.h>
+
+#include "lex/lex.h"
+#include "utils/parserutilserror.h"
+
+/** \todo Optimisation -- we're currently revisiting a bunch of input 
+ *	  characters (Currently, we're calling parserutils_inputstream_peek about
+ *	  1.5x the number of characters in the input stream). Ideally, 
+ *	  we'll visit each character in the input exactly once. In reality, 
+ *	  the upper bound is twice, due to the need, in some cases, to read 
+ *	  one character beyond the end of a token's input to detect the end 
+ *	  of the token. Resumability adds a little overhead here, unless 
+ *	  we're somewhat more clever when it comes to having support for 
+ *	  restarting mid-escape sequence. Currently, we rewind back to the 
+ *	  start of the sequence and process the whole thing again.
+ */
+
+enum {
+	sSTART		=  0,
+	sATKEYWORD	=  1,
+	sSTRING		=  2,
+	sHASH		=  3,
+	sNUMBER		=  4, 
+	sCDO		=  5,
+	sCDC		=  6,
+	sS		=  7,
+	sCOMMENT	=  8,
+	sMATCH		=  9, 
+	sURI		= 10,
+	sIDENT		= 11,
+	sESCAPEDIDENT	= 12,
+	sURL		= 13,
+	sUCR		= 14 
+};
+
+/**
+ * CSS lexer object
+ */
+struct css_lexer
+{
+	parserutils_inputstream *input;	/**< Inputstream containing CSS */
+
+	size_t bytesReadForToken;	/**< Total bytes read from the 
+					 * inputstream for the current token */
+
+	css_token token;		/**< The current token */
+
+	bool escapeSeen;		/**< Whether an escape sequence has 
+					 * been seen while processing the input
+					 * for the current token */
+	parserutils_buffer *unescapedTokenData;	/**< Buffer containing 
+					 	 * unescaped token data 
+						 * (used iff escapeSeen == true)
+						 */
+
+	uint32_t state    : 4,		/**< Current state */
+		 substate : 4;		/**< Current substate */
+
+	struct {
+		uint8_t first;		/**< First character read for token */
+		size_t origBytes;	/**< Storage of current number of 
+					 * bytes read, for rewinding */
+		bool lastWasStar;	/**< Whether the previous character 
+					 * was an asterisk */
+		bool lastWasCR;		/**< Whether the previous character
+					 * was CR */
+		size_t bytesForURL;	/**< Input bytes read for "url(", for 
+					 * rewinding */
+		size_t dataLenForURL;	/**< Output length for "url(", for
+					 * rewinding */
+		int hexCount;		/**< Counter for reading hex digits */
+	} context;			/**< Context for the current state */
+
+	bool emit_comments;		/**< Whether to emit comment tokens */
+
+	uint32_t currentCol;		/**< Current column in source */
+	uint32_t currentLine;		/**< Current line in source */
+
+	css_alloc alloc;		/**< Memory (de)allocation function */
+	void *pw;			/**< Pointer to client-specific data */
+};
+
+#define APPEND(lexer, data, len)					\
+do {									\
+	css_error error;						\
+	error = appendToTokenData((lexer), 				\
+			(const uint8_t*) (data), (len));		\
+	if (error != CSS_OK)						\
+		return error;						\
+	(lexer)->bytesReadForToken += (len);				\
+	(lexer)->currentCol += (len);					\
+} while(0)								\
+
+static inline css_error appendToTokenData(css_lexer *lexer, 
+		const uint8_t *data, size_t len);
+static inline css_error emitToken(css_lexer *lexer, css_token_type type,
+		const css_token **token);
+
+static inline css_error AtKeyword(css_lexer *lexer, const css_token **token);
+static inline css_error CDCOrIdentOrFunction(css_lexer *lexer,
+		const css_token **token);
+static inline css_error CDO(css_lexer *lexer, const css_token **token);
+static inline css_error Comment(css_lexer *lexer, const css_token **token);
+static inline css_error EscapedIdentOrFunction(css_lexer *lexer,
+		const css_token **token);
+static inline css_error Hash(css_lexer *lexer, const css_token **token);
+static inline css_error IdentOrFunction(css_lexer *lexer,
+		const css_token **token);
+static inline css_error Match(css_lexer *lexer, const css_token **token);
+static inline css_error NumberOrPercentageOrDimension(css_lexer *lexer,
+		const css_token **token);
+static inline css_error S(css_lexer *lexer, const css_token **token);
+static inline css_error Start(css_lexer *lexer, const css_token **token);
+static inline css_error String(css_lexer *lexer, const css_token **token);
+static inline css_error URIOrUnicodeRangeOrIdentOrFunction(
+		css_lexer *lexer, const css_token **token);
+static inline css_error URI(css_lexer *lexer, const css_token **token);
+static inline css_error UnicodeRange(css_lexer *lexer, const css_token **token);
+
+static inline css_error consumeDigits(css_lexer *lexer);
+static inline css_error consumeEscape(css_lexer *lexer, bool nl);
+static inline css_error consumeNMChars(css_lexer *lexer);
+static inline css_error consumeString(css_lexer *lexer);
+static inline css_error consumeStringChars(css_lexer *lexer);
+static inline css_error consumeUnicode(css_lexer *lexer, uint32_t ucs);
+static inline css_error consumeURLChars(css_lexer *lexer);
+static inline css_error consumeWChars(css_lexer *lexer);
+
+static inline uint32_t charToHex(uint8_t c);
+static inline bool startNMChar(uint8_t c);
+static inline bool startNMStart(uint8_t c);
+static inline bool startStringChar(uint8_t c);
+static inline bool startURLChar(uint8_t c);
+static inline bool isDigit(uint8_t c);
+static inline bool isHex(uint8_t c);
+static inline bool isSpace(uint8_t c);
+
+/**
+ * Create a lexer instance
+ *
+ * \param input  The inputstream to read from
+ * \param alloc  Memory (de)allocation function
+ * \param pw     Pointer to client-specific private data
+ * \return Pointer to instance, or NULL on memory exhaustion
+ */
+css_lexer *css_lexer_create(parserutils_inputstream *input, 
+		css_alloc alloc, void *pw)
+{
+	css_lexer *lex;
+
+	if (input == NULL || alloc == NULL)
+		return NULL;
+
+	lex = alloc(NULL, sizeof(css_lexer), pw);
+	if (lex == NULL)
+		return NULL;
+
+	lex->input = input;
+	lex->bytesReadForToken = 0;
+	lex->token.type = CSS_TOKEN_EOF;
+	lex->token.data.ptr = NULL;
+	lex->token.data.len = 0;
+	lex->escapeSeen = false;
+	lex->unescapedTokenData = NULL;
+	lex->state = sSTART;
+	lex->substate = 0;
+	lex->emit_comments = false;
+	lex->currentCol = 1;
+	lex->currentLine = 1;
+	lex->alloc = alloc;
+	lex->pw = pw;
+
+	return lex;
+}
+
+/**
+ * Destroy a lexer instance
+ *
+ * \param lexer  The instance to destroy
+ */
+void css_lexer_destroy(css_lexer *lexer)
+{
+	if (lexer == NULL)
+		return;
+
+	if (lexer->unescapedTokenData != NULL)
+		parserutils_buffer_destroy(lexer->unescapedTokenData);
+
+	lexer->alloc(lexer, 0, lexer->pw);
+}
+
+/**
+ * Configure a lexer instance
+ *
+ * \param lexer   The lexer to configure
+ * \param type    The option type to modify
+ * \param params  Option-specific parameters
+ * \return CSS_OK on success, appropriate error otherwise
+ */
+css_error css_lexer_setopt(css_lexer *lexer, css_lexer_opttype type,
+		css_lexer_optparams *params)
+{
+	if (lexer == NULL || params == NULL)
+		return CSS_BADPARM;
+
+	switch (type) {
+	case CSS_LEXER_EMIT_COMMENTS:
+		lexer->emit_comments = params->emit_comments;
+		break;
+	default:
+		return CSS_BADPARM;
+	}
+
+	return CSS_OK;
+}
+
+/**
+ * Retrieve a token from a lexer
+ *
+ * \param lexer  The lexer instance to read from
+ * \param token  Pointer to location to receive pointer to token
+ * \return CSS_OK on success, appropriate error otherwise
+ */
+css_error css_lexer_get_token(css_lexer *lexer, const css_token **token)
+{
+	css_error error;
+
+	if (lexer == NULL || token == NULL)
+		return CSS_BADPARM;
+
+	switch (lexer->state)
+	{
+	case sSTART:
+	start:
+		return Start(lexer, token);
+	case sATKEYWORD:
+		return AtKeyword(lexer, token);
+	case sSTRING:
+		return String(lexer, token);
+	case sHASH:
+		return Hash(lexer, token);
+	case sNUMBER:
+		return NumberOrPercentageOrDimension(lexer, token);
+	case sCDO:
+		return CDO(lexer, token);
+	case sCDC:
+		return CDCOrIdentOrFunction(lexer, token);
+	case sS:
+		return S(lexer, token);
+	case sCOMMENT:
+		error = Comment(lexer, token);
+		if (!lexer->emit_comments && error == CSS_OK)
+			goto start;
+		return error;
+	case sMATCH:
+		return Match(lexer, token);
+	case sURI:
+		return URI(lexer, token);
+	case sIDENT:
+		return IdentOrFunction(lexer, token);
+	case sESCAPEDIDENT:
+		return EscapedIdentOrFunction(lexer, token);
+	case sURL:
+		return URI(lexer, token);
+	case sUCR:
+		return UnicodeRange(lexer, token);
+	}
+
+	/* Should never be reached */
+	assert(0);
+
+	return CSS_OK;
+}
+
+/******************************************************************************
+ * Utility routines                                                           *
+ ******************************************************************************/
+
+/**
+ * Append some data to the current token
+ *
+ * \param lexer  The lexer instance
+ * \param data   Pointer to data to append
+ * \param len    Length, in bytes, of data
+ * \return CSS_OK on success, appropriate error otherwise
+ *
+ * This should not be called directly without good reason. Use the APPEND()
+ * macro instead. 
+ */
+css_error appendToTokenData(css_lexer *lexer, const uint8_t *data, size_t len)
+{
+	css_token *token = &lexer->token;
+
+	if (lexer->escapeSeen) {
+		css_error error = css_error_from_parserutils_error(
+				parserutils_buffer_append(
+					lexer->unescapedTokenData, data, len));
+		if (error != CSS_OK)
+			return error;
+	}
+
+	token->data.len += len;
+
+	return CSS_OK;
+}
+
+/**
+ * Prepare a token for consumption and emit it to the client
+ *
+ * \param lexer  The lexer instance
+ * \param type   The type of token to emit
+ * \param token  Pointer to location to receive pointer to token
+ * \return CSS_OK on success, appropriate error otherwise
+ */
+css_error emitToken(css_lexer *lexer, css_token_type type,
+		const css_token **token)
+{
+	css_token *t = &lexer->token;
+
+	t->type = type;
+
+	/* Calculate token data start pointer. We have to do this here as 
+	 * the inputstream's buffer may have moved under us. */
+	if (lexer->escapeSeen) {
+		t->data.ptr = lexer->unescapedTokenData->data;
+	} else {
+		size_t clen;
+		uintptr_t ptr = parserutils_inputstream_peek(
+				lexer->input, 0, &clen);
+
+		assert(type == CSS_TOKEN_EOF || 
+				(ptr != PARSERUTILS_INPUTSTREAM_EOF && 
+				ptr != PARSERUTILS_INPUTSTREAM_OOD));
+
+		t->data.ptr = (type == CSS_TOKEN_EOF) ? NULL : (uint8_t *) ptr;
+	}
+
+	switch (type) {
+	case CSS_TOKEN_ATKEYWORD:
+		/* Strip the '@' from the front */
+		t->data.ptr += 1;
+		t->data.len -= 1;
+		break;
+	case CSS_TOKEN_STRING:
+		/* Strip the leading quote */
+		t->data.ptr += 1;
+		t->data.len -= 1;
+
+		/* Strip the trailing quote */
+		t->data.len -= 1;
+		break;
+	case CSS_TOKEN_HASH:
+		/* Strip the '#' from the front */
+		t->data.ptr += 1;
+		t->data.len -= 1;
+		break;
+	case CSS_TOKEN_PERCENTAGE:
+		/* Strip the '%' from the end */
+		t->data.len -= 1;
+		break;
+	case CSS_TOKEN_DIMENSION:
+		/** \todo Do we want to separate the value from the units? */
+		break;
+	case CSS_TOKEN_URI:
+		/* Strip the "url(" from the start */
+		t->data.ptr += sizeof("url(") - 1;
+		t->data.len -= sizeof("url(") - 1;
+
+		/* Strip any leading whitespace */
+		while (isSpace(t->data.ptr[0])) {
+			t->data.ptr++;
+			t->data.len--;
+		}
+
+		/* Strip any leading quote */
+		if (t->data.ptr[0] == '"' || t->data.ptr[0] == '\'') {
+			t->data.ptr += 1;
+			t->data.len -= 1;
+		}
+
+		/* Strip the trailing ')' */
+		t->data.len -= 1;
+
+		/* Strip any trailing whitespace */
+		while (isSpace(t->data.ptr[t->data.len - 1])) {
+			t->data.len--;
+		}
+
+		/* Strip any trailing quote */
+		if (t->data.ptr[t->data.len - 1] == '"' || 
+				t->data.ptr[t->data.len - 1] == '\'') {
+			t->data.len -= 1;
+		}
+		break;
+	case CSS_TOKEN_UNICODE_RANGE:
+		/* Remove "U+" from the start */
+		t->data.ptr += sizeof("U+") - 1;
+		t->data.len -= sizeof("U+") - 1;
+		break;
+	case CSS_TOKEN_COMMENT:
+		/* Strip the leading '/' and '*' */
+		t->data.ptr += sizeof("/*") - 1;
+		t->data.len -= sizeof("/*") - 1;
+
+		/* Strip the trailing '*' and '/' */
+		t->data.len -= sizeof("*/") - 1;
+		break;
+	case CSS_TOKEN_FUNCTION:
+		/* Strip the trailing '(' */
+		t->data.len -= 1;
+		break;
+	default:
+		break;
+	}
+
+	*token = t;
+
+	/* Reset the lexer's state */
+	lexer->state = sSTART;
+	lexer->substate = 0;
+
+	return CSS_OK;
+}
+
+/******************************************************************************
+ * State machine components                                                   *
+ ******************************************************************************/
+
+css_error AtKeyword(css_lexer *lexer, const css_token **token)
+{
+	uintptr_t cptr;
+	uint8_t c;
+	size_t clen;
+	css_error error;
+	enum { Initial = 0, Escape = 1, NMChar = 2 };
+
+	/* ATKEYWORD = '@' ident 
+	 * 
+	 * The '@' has been consumed.
+	 */
+
+	switch (lexer->substate) {
+	case Initial:
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF)
+			return emitToken(lexer, CSS_TOKEN_CHAR, token);
+
+		c = *((uint8_t *) cptr);
+
+		if (!startNMStart(c))
+			return emitToken(lexer, CSS_TOKEN_CHAR, token);
+
+		if (c != '\\') {
+			APPEND(lexer, cptr, clen);
+		} else {
+			lexer->bytesReadForToken += clen;
+			goto escape;
+		}
+
+		/* Fall through */
+	case NMChar:
+	nmchar:
+		lexer->substate = NMChar;
+		error = consumeNMChars(lexer);
+		if (error != CSS_OK)
+			return error;
+		break;
+
+	case Escape:
+	escape:
+		lexer->substate = Escape;
+		error = consumeEscape(lexer, false);
+		if (error != CSS_OK) {
+			if (error == CSS_EOF || error == CSS_INVALID) {
+				/* Rewind the '\\' */
+				lexer->bytesReadForToken -= 1;
+
+				return emitToken(lexer, CSS_TOKEN_CHAR, token);
+			}
+
+			return error;
+		}
+
+		goto nmchar;
+	}
+
+	return emitToken(lexer, CSS_TOKEN_ATKEYWORD, token);
+}
+
+css_error CDCOrIdentOrFunction(css_lexer *lexer, const css_token **token)
+{
+	css_token *t = &lexer->token;
+	uintptr_t cptr;
+	uint8_t c;
+	size_t clen;
+	css_error error;
+	enum { Initial = 0, Escape = 1, Gt = 2 };
+
+	/* CDC = "-->"
+	 * IDENT = [-]? nmstart nmchar*
+	 * FUNCTION = [-]? nmstart nmchar* '('
+	 *
+	 * The first dash has been consumed. Thus, we must consume the next 
+	 * character in the stream. If it's a dash, then we're dealing with 
+	 * CDC. Otherwise, we're dealing with IDENT/FUNCTION.
+	 */
+
+	switch (lexer->substate) {
+	case Initial:
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF) {
+			/* We can only match char with what we've read so far */
+			return emitToken(lexer, CSS_TOKEN_CHAR, token);
+		}
+
+		c = *((uint8_t *) cptr);
+
+		if (c != '-' && !startNMStart(c)) {
+			/* Can only be CHAR */
+			return emitToken(lexer, CSS_TOKEN_CHAR, token);
+		}
+
+
+		if (c != '\\') {
+			APPEND(lexer, cptr, clen);
+		}
+
+		if (c != '-') {
+			if (c == '\\') {
+				lexer->bytesReadForToken += clen;
+				goto escape;
+			}
+
+			lexer->state = sIDENT;
+			lexer->substate = 0;
+			return IdentOrFunction(lexer, token);
+		}
+
+		/* Fall through */
+	case Gt:
+		lexer->substate = Gt;
+
+		/* Ok, so we're dealing with CDC. Expect a '>' */
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF) {
+			/* CHAR is the only match here */
+			/* Remove the '-' we read above */
+			lexer->bytesReadForToken -= 1;
+			t->data.len -= 1;
+			return emitToken(lexer, CSS_TOKEN_CHAR, token);
+		}
+
+		c = *((uint8_t *) cptr);
+
+		if (c == '>') {
+			APPEND(lexer, cptr, clen);
+
+			t->type = CSS_TOKEN_CDC;
+		} else {
+			/* Remove the '-' we read above */
+			lexer->bytesReadForToken -= 1;
+			t->data.len -= 1;
+			t->type = CSS_TOKEN_CHAR;
+		}
+		break;
+
+	case Escape:
+	escape:
+		lexer->substate = Escape;
+		error = consumeEscape(lexer, false);
+		if (error != CSS_OK) {
+			if (error == CSS_EOF || error == CSS_INVALID) {
+				/* Rewind the '\\' */
+				lexer->bytesReadForToken -= 1;
+
+				return emitToken(lexer, CSS_TOKEN_CHAR, token);
+			}
+
+			return error;
+		}
+
+		lexer->state = sIDENT;
+		lexer->substate = 0;
+		return IdentOrFunction(lexer, token);
+	}
+
+	return emitToken(lexer, t->type, token);
+}
+
+css_error CDO(css_lexer *lexer, const css_token **token)
+{
+	css_token *t = &lexer->token;
+	uintptr_t cptr;
+	uint8_t c;
+	size_t clen;
+	enum { Initial = 0, Dash1 = 1, Dash2 = 2 };
+
+	/* CDO = "<!--"
+	 * 
+	 * The '<' has been consumed
+	 */
+
+	switch (lexer->substate) {
+	case Initial:
+		/* Expect '!' */
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF) {
+			/* CHAR is the only match here */
+			return emitToken(lexer, CSS_TOKEN_CHAR, token);
+		}
+
+		c = *((uint8_t *) cptr);
+
+		if (c == '!') {
+			APPEND(lexer, cptr, clen);
+		} else {
+			return emitToken(lexer, CSS_TOKEN_CHAR, token);
+		}
+
+		/* Fall Through */
+	case Dash1:
+		lexer->substate = Dash1;
+
+		/* Expect '-' */
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF) {
+			/* CHAR is the only match here */
+			/* Remove the '!' we read above */
+			lexer->bytesReadForToken -= 1;
+			t->data.len -= 1;
+			return emitToken(lexer, CSS_TOKEN_CHAR, token);
+		}
+
+		c = *((uint8_t *) cptr);
+
+		if (c == '-') {
+			APPEND(lexer, cptr, clen);
+		} else {
+			/* Remove the '!' we read above */
+			lexer->bytesReadForToken -= 1;
+			t->data.len -= 1;
+			return emitToken(lexer, CSS_TOKEN_CHAR, token);
+		}
+
+		/* Fall through */
+	case Dash2:
+		lexer->substate = Dash2;
+
+		/* Expect '-' */
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF) {
+			/* CHAR is the only match here */
+			/* Remove the '-' and the '!' we read above */
+			lexer->bytesReadForToken -= 2;
+			t->data.len -= 2;
+			return emitToken(lexer, CSS_TOKEN_CHAR, token);
+		}
+
+		c = *((uint8_t *) cptr);
+
+		if (c == '-') {
+			APPEND(lexer, cptr, clen);
+		} else {
+			/* Remove the '-' and the '!' we read above */
+			lexer->bytesReadForToken -= 2;
+			t->data.len -= 2;
+			return emitToken(lexer, CSS_TOKEN_CHAR, token);
+		}
+	}
+
+	return emitToken(lexer, CSS_TOKEN_CDO, token);
+}
+
+css_error Comment(css_lexer *lexer, const css_token **token)
+{
+	uintptr_t cptr;
+	uint8_t c;
+	size_t clen;
+	enum { Initial = 0, InComment = 1 };
+
+	/* COMMENT = '/' '*' [^*]* '*'+ ([^/] [^*]* '*'+)* '/'
+	 *
+	 * The '/' has been consumed.
+	 */
+	switch (lexer->substate) {
+	case Initial:
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF)
+			return emitToken(lexer, CSS_TOKEN_CHAR, token);
+
+		c = *((uint8_t *) cptr);
+
+		if (c != '*')
+			return emitToken(lexer, CSS_TOKEN_CHAR, token);
+
+		APPEND(lexer, cptr, clen);
+	
+		/* Fall through */
+	case InComment:
+		lexer->substate = InComment;
+
+		while (1) {
+			cptr = parserutils_inputstream_peek(lexer->input,
+					lexer->bytesReadForToken, &clen);
+			if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+				return CSS_NEEDDATA;
+
+			if (cptr == PARSERUTILS_INPUTSTREAM_EOF) {
+				/* As per unterminated strings, 
+				 * we ignore unterminated comments. */
+				return emitToken(lexer, CSS_TOKEN_EOF, token);
+			}
+
+			c = *((uint8_t *) cptr);
+
+			APPEND(lexer, cptr, clen);
+
+			if (lexer->context.lastWasStar && c == '/')
+				break;
+
+			lexer->context.lastWasStar = (c == '*');
+
+			if (c == '\n' || c == '\f') {
+				lexer->currentCol = 1;
+				lexer->currentLine++;
+			}
+
+			if (lexer->context.lastWasCR && c != '\n') {
+				lexer->currentCol = 1;
+				lexer->currentLine++;
+			}
+			lexer->context.lastWasCR = (c == '\r');
+		}
+	}
+
+	return emitToken(lexer, CSS_TOKEN_COMMENT, token);
+}
+
+css_error EscapedIdentOrFunction(css_lexer *lexer, const css_token **token)
+{
+	css_error error;
+
+	/* IDENT = ident = [-]? nmstart nmchar*
+	 * FUNCTION = ident '(' = [-]? nmstart nmchar* '('
+	 *
+	 * In this case, nmstart is an escape sequence and no '-' is present.
+	 *
+	 * The '\\' has been consumed.
+	 */
+
+	error = consumeEscape(lexer, false);
+	if (error != CSS_OK) {
+		if (error == CSS_EOF || error == CSS_INVALID) {
+			/* The '\\' is a token of its own */
+			return emitToken(lexer, CSS_TOKEN_CHAR, token);
+		}
+
+		return error;
+	}
+
+	lexer->state = sIDENT;
+	lexer->substate = 0;
+	return IdentOrFunction(lexer, token);
+}
+
+css_error Hash(css_lexer *lexer, const css_token **token)
+{	
+	css_error error;
+	
+	/* HASH = '#' name  = '#' nmchar+ 
+	 *
+	 * The '#' has been consumed.
+	 */
+
+	error = consumeNMChars(lexer);
+	if (error != CSS_OK)
+		return error;
+
+	/* Require at least one NMChar otherwise, we're just a raw '#' */
+	if (lexer->bytesReadForToken - lexer->context.origBytes > 0)
+		return emitToken(lexer, CSS_TOKEN_HASH, token);
+
+	return emitToken(lexer, CSS_TOKEN_CHAR, token);
+}
+
+css_error IdentOrFunction(css_lexer *lexer, const css_token **token)
+{
+	css_token *t = &lexer->token;
+	uintptr_t cptr;
+	uint8_t c;
+	size_t clen;
+	css_error error;
+	enum { Initial = 0, Bracket = 1 };
+
+	/* IDENT = ident = [-]? nmstart nmchar*
+	 * FUNCTION = ident '(' = [-]? nmstart nmchar* '('
+	 *
+	 * The optional dash and nmstart are already consumed
+	 */
+
+	switch (lexer->substate) {
+	case Initial:
+		/* Consume all subsequent nmchars (if any exist) */
+		error = consumeNMChars(lexer);
+		if (error != CSS_OK)
+			return error;
+
+		/* Fall through */
+	case Bracket:
+		lexer->substate = Bracket;
+
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF) {
+			/* IDENT, rather than CHAR */
+			return emitToken(lexer, CSS_TOKEN_IDENT, token);
+		}
+
+		c = *((uint8_t *) cptr);
+
+		if (c == '(') {
+			APPEND(lexer, cptr, clen);
+
+			t->type = CSS_TOKEN_FUNCTION;
+		} else {
+			t->type = CSS_TOKEN_IDENT;
+		}
+	}
+
+	return emitToken(lexer, t->type, token);
+}
+
+css_error Match(css_lexer *lexer, const css_token **token)
+{
+	uintptr_t cptr;
+	uint8_t c;
+	size_t clen;
+	css_token_type type = CSS_TOKEN_EOF; /* GCC's braindead */
+
+	/* INCLUDES       = "~="
+	 * DASHMATCH      = "|="
+	 * PREFIXMATCH    = "^="
+	 * SUFFIXMATCH    = "$="
+	 * SUBSTRINGMATCH = "*="
+	 *
+	 * The first character has been consumed.
+	 */
+
+	cptr = parserutils_inputstream_peek(lexer->input,
+			lexer->bytesReadForToken, &clen);
+	if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+		return CSS_NEEDDATA;
+
+	if (cptr == PARSERUTILS_INPUTSTREAM_EOF)
+		return emitToken(lexer, CSS_TOKEN_CHAR, token);
+
+	c = *((uint8_t *) cptr);
+
+	if (c != '=')
+		return emitToken(lexer, CSS_TOKEN_CHAR, token);
+
+	APPEND(lexer, cptr, clen);
+
+	switch (lexer->context.first) {
+	case '~':
+		type = CSS_TOKEN_INCLUDES;
+		break;
+	case '|':
+		type = CSS_TOKEN_DASHMATCH;
+		break;
+	case '^':
+		type = CSS_TOKEN_PREFIXMATCH;
+		break;
+	case '$':
+		type = CSS_TOKEN_SUFFIXMATCH;
+		break;
+	case '*':
+		type = CSS_TOKEN_SUBSTRINGMATCH;
+		break;
+	default:
+		assert(0);
+	}
+
+	return emitToken(lexer, type, token);
+}
+
+css_error NumberOrPercentageOrDimension(css_lexer *lexer, 
+		const css_token **token)
+{
+	css_token *t = &lexer->token;
+	uintptr_t cptr;
+	uint8_t c;
+	size_t clen;
+	css_error error;
+	enum { Initial = 0, Dot = 1, MoreDigits = 2, 
+		Suffix = 3, NMChars = 4, Escape = 5 };
+
+	/* NUMBER = num = [0-9]+ | [0-9]* '.' [0-9]+
+	 * PERCENTAGE = num '%'
+	 * DIMENSION = num ident
+	 *
+	 * The first digit, or '.' has been consumed.
+	 */
+
+	switch (lexer->substate) {
+	case Initial:
+		error = consumeDigits(lexer);
+		if (error != CSS_OK)
+			return error;
+
+		/* Fall through */
+	case Dot:
+		lexer->substate = Dot;
+
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF) {
+			if (t->data.len == 1 && lexer->context.first == '.')
+				return emitToken(lexer, CSS_TOKEN_CHAR, token);
+			else
+				return emitToken(lexer, CSS_TOKEN_NUMBER, 
+						token);
+		}
+
+		c = *((uint8_t *) cptr);
+
+		/* Bail if we've not got a '.' or we've seen one already */
+		if (c != '.' || lexer->context.first == '.')
+			goto suffix;
+
+		/* Save the token length up to the end of the digits */
+		lexer->context.origBytes = lexer->bytesReadForToken;
+
+		/* Append the '.' to the token */
+		APPEND(lexer, cptr, clen);
+
+		/* Fall through */
+	case MoreDigits:
+		lexer->substate = MoreDigits;
+
+		error = consumeDigits(lexer);
+		if (error != CSS_OK)
+			return error;
+
+		if (lexer->bytesReadForToken - lexer->context.origBytes == 1) {
+			/* No digits after dot => dot isn't part of number */
+			lexer->bytesReadForToken -= 1;
+			t->data.len -= 1;
+		}
+
+		/* Fall through */
+	case Suffix:
+	suffix:
+		lexer->substate = Suffix;
+
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF) {
+			if (t->data.len == 1 && lexer->context.first == '.')
+				return emitToken(lexer, CSS_TOKEN_CHAR, token);
+			else
+				return emitToken(lexer, CSS_TOKEN_NUMBER, 
+						token);
+		}
+
+		c = *((uint8_t *) cptr);
+
+		/* A solitary '.' is a CHAR, not numeric */
+		if (t->data.len == 1 && lexer->context.first == '.')
+			return emitToken(lexer, CSS_TOKEN_CHAR, token);
+
+		if (c == '%') {
+			APPEND(lexer, cptr, clen);
+
+			return emitToken(lexer, CSS_TOKEN_PERCENTAGE, token);
+		} else if (!startNMStart(c)) {
+			return emitToken(lexer, CSS_TOKEN_NUMBER, token);
+		}
+
+		if (c != '\\') {
+			APPEND(lexer, cptr, clen);
+		} else {
+			lexer->bytesReadForToken += clen;
+			goto escape;
+		}
+
+		/* Fall through */
+	case NMChars:
+	nmchars:
+		lexer->substate = NMChars;
+
+		error = consumeNMChars(lexer);
+		if (error != CSS_OK)
+			return error;
+
+		break;
+	case Escape:
+	escape:
+		lexer->substate = Escape;
+
+		error = consumeEscape(lexer, false);
+		if (error != CSS_OK) {
+			if (error == CSS_EOF || error == CSS_INVALID) {
+				/* Rewind the '\\' */
+				lexer->bytesReadForToken -= 1;
+
+				/* This can only be a number */
+				return emitToken(lexer, 
+						CSS_TOKEN_NUMBER, token);
+			}
+
+			return error;
+		}
+
+		goto nmchars;
+	}
+
+	return emitToken(lexer, CSS_TOKEN_DIMENSION, token);
+}
+
+css_error S(css_lexer *lexer, const css_token **token)
+{
+	css_error error;
+
+	/* S = wc*
+	 * 
+	 * The first whitespace character has been consumed.
+	 */
+
+	error = consumeWChars(lexer);
+	if (error != CSS_OK)
+		return error;
+
+	return emitToken(lexer, CSS_TOKEN_S, token);
+}
+
+css_error Start(css_lexer *lexer, const css_token **token)
+{
+	css_token *t = &lexer->token;
+	uintptr_t cptr;
+	uint8_t c;
+	size_t clen;
+	css_error error;
+
+start:
+
+	/* Advance past the input read for the previous token */
+	if (lexer->bytesReadForToken > 0) {
+		parserutils_inputstream_advance(
+				lexer->input, lexer->bytesReadForToken);
+		lexer->bytesReadForToken = 0;
+	}
+
+	/* Reset in preparation for the next token */
+	t->type = CSS_TOKEN_EOF;
+	t->data.ptr = NULL;
+	t->data.len = 0;
+	t->col = lexer->currentCol;
+	t->line = lexer->currentLine;
+	lexer->escapeSeen = false;
+	if (lexer->unescapedTokenData != NULL)
+		lexer->unescapedTokenData->length = 0;
+
+	cptr = parserutils_inputstream_peek(lexer->input, 0, &clen);
+	if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+		return CSS_NEEDDATA;
+
+	if (cptr == PARSERUTILS_INPUTSTREAM_EOF)
+		return emitToken(lexer, CSS_TOKEN_EOF, token);
+
+	APPEND(lexer, cptr, clen);
+
+	c = *((uint8_t *) cptr);
+
+	if (clen > 1 || c >= 0x80) {
+		lexer->state = sIDENT;
+		lexer->substate = 0;
+		return IdentOrFunction(lexer, token);
+	}
+
+	switch (c) {
+	case '@':
+		lexer->state = sATKEYWORD;
+		lexer->substate = 0;
+		return AtKeyword(lexer, token);
+	case '"': case '\'':
+		lexer->state = sSTRING;
+		lexer->substate = 0;
+		lexer->context.first = c;
+		return String(lexer, token);
+	case '#':
+		lexer->state = sHASH;
+		lexer->substate = 0;
+		lexer->context.origBytes = lexer->bytesReadForToken;
+		return Hash(lexer, token);
+	case '0': case '1': case '2': case '3': case '4': 
+	case '5': case '6': case '7': case '8': case '9':
+	case '.':
+		lexer->state = sNUMBER;
+		lexer->substate = 0;
+		lexer->context.first = c;
+		return NumberOrPercentageOrDimension(lexer, token);
+	case '<':
+		lexer->state = sCDO;
+		lexer->substate = 0;
+		return CDO(lexer, token);
+	case '-':
+		lexer->state = sCDC;
+		lexer->substate = 0;
+		return CDCOrIdentOrFunction(lexer, token);
+	case ' ': case '\t': case '\r': case '\n': case '\f':
+		lexer->state = sS;
+		lexer->substate = 0;
+		if (c == '\n' || c == '\f') {
+			lexer->currentCol = 1;
+			lexer->currentLine++;
+		}
+		lexer->context.lastWasCR = (c == '\r');
+		return S(lexer, token);
+	case '/':
+		lexer->state = sCOMMENT;
+		lexer->substate = 0;
+		lexer->context.lastWasStar = false;
+		lexer->context.lastWasCR = false;
+		error = Comment(lexer, token);
+		if (!lexer->emit_comments && error == CSS_OK)
+			goto start;
+		return error;
+	case '~': case '|': case '^': case '$': case '*':
+		lexer->state = sMATCH;
+		lexer->substate = 0;
+		lexer->context.first = c;
+		return Match(lexer, token);
+	case 'u': case 'U':
+		lexer->state = sURI;
+		lexer->substate = 0;
+		return URIOrUnicodeRangeOrIdentOrFunction(lexer, token);
+	case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': 
+	case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': 
+	case 'm': case 'n': case 'o': case 'p': case 'q': case 'r':
+	case 's': case 't': /*  'u'*/ case 'v': case 'w': case 'x': 
+	case 'y': case 'z':
+	case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
+	case 'G': case 'H': case 'I': case 'J': case 'K': case 'L':
+	case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R':
+	case 'S': case 'T': /*  'U'*/ case 'V': case 'W': case 'X':
+	case 'Y': case 'Z':
+	case '_': 
+		lexer->state = sIDENT;
+		lexer->substate = 0;
+		return IdentOrFunction(lexer, token);
+	case '\\':
+		lexer->state = sESCAPEDIDENT;
+		lexer->substate = 0;
+		return EscapedIdentOrFunction(lexer, token);
+	default:
+		return emitToken(lexer, CSS_TOKEN_CHAR, token);
+	}
+}
+
+css_error String(css_lexer *lexer, const css_token **token)
+{
+	css_error error;
+
+	/* STRING = string
+	 *
+	 * The open quote has been consumed.
+	 */
+
+	error = consumeString(lexer);
+	if (error != CSS_OK && error != CSS_EOF)
+		return error;
+
+	return emitToken(lexer, 
+			error == CSS_EOF ? CSS_TOKEN_EOF : CSS_TOKEN_STRING, 
+			token);
+}
+
+css_error URIOrUnicodeRangeOrIdentOrFunction(css_lexer *lexer, 
+		const css_token **token)
+{
+	uintptr_t cptr;
+	uint8_t c;
+	size_t clen;
+
+	/* URI = "url(" w (string | urlchar*) w ')' 
+	 * UNICODE-RANGE = [Uu] '+' [0-9a-fA-F?]{1,6}(-[0-9a-fA-F]{1,6})?
+	 * IDENT = ident = [-]? nmstart nmchar*
+	 * FUNCTION = ident '(' = [-]? nmstart nmchar* '('
+	 *
+	 * The 'u' (or 'U') has been consumed.
+	 */
+
+	cptr = parserutils_inputstream_peek(lexer->input, 
+			lexer->bytesReadForToken, &clen);
+	if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+		return CSS_NEEDDATA;
+
+	if (cptr == PARSERUTILS_INPUTSTREAM_EOF) {
+		/* IDENT, rather than CHAR */
+		return emitToken(lexer, CSS_TOKEN_IDENT, token);
+	}
+
+	c = *((uint8_t *) cptr);
+
+	if (c == 'r' || c == 'R') {
+		APPEND(lexer, cptr, clen);
+
+		lexer->state = sURL;
+		lexer->substate = 0;
+		return URI(lexer, token);
+	} else if (c == '+') {
+		APPEND(lexer, cptr, clen);
+
+		lexer->state = sUCR;
+		lexer->substate = 0;
+		lexer->context.hexCount = 0;
+		return UnicodeRange(lexer, token);
+	}
+
+	/* Can only be IDENT or FUNCTION. Reprocess current character */
+	lexer->state = sIDENT;
+	lexer->substate = 0;
+	return IdentOrFunction(lexer, token);
+}
+
+css_error URI(css_lexer *lexer, const css_token **token)
+{
+	uintptr_t cptr;
+	uint8_t c;
+	size_t clen;
+	css_error error;
+	enum { Initial = 0, LParen = 1, W1 = 2, Quote = 3, 
+		URL = 4, W2 = 5, RParen = 6, String = 7 };
+
+	/* URI = "url(" w (string | urlchar*) w ')' 
+	 *
+	 * 'u' and 'r' have been consumed.
+	 */
+
+	switch (lexer->substate) {
+	case Initial:
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF) {
+			/* IDENT */
+			return emitToken(lexer, CSS_TOKEN_IDENT, token);
+		}
+
+		c = *((uint8_t *) cptr);
+
+		if (c == 'l' || c == 'L') {
+			APPEND(lexer, cptr, clen);
+		} else {
+			lexer->state = sIDENT;
+			lexer->substate = 0;
+			return IdentOrFunction(lexer, token);
+		}
+
+		/* Fall through */
+	case LParen:
+		lexer->substate = LParen;
+
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF)
+			return emitToken(lexer, CSS_TOKEN_IDENT, token);
+
+		c = *((uint8_t *) cptr);
+
+		if (c == '(') {
+			APPEND(lexer, cptr, clen);
+		} else {
+			lexer->state = sIDENT;
+			lexer->substate = 0;
+			return IdentOrFunction(lexer, token);
+		}
+
+		/* Save the number of input bytes read for "url(" */
+		lexer->context.bytesForURL = lexer->bytesReadForToken;
+		/* And the length of the token data at the same point */
+		lexer->context.dataLenForURL = lexer->token.data.len;
+
+		lexer->context.lastWasCR = false;
+
+		/* Fall through */
+	case W1:
+		lexer->substate = W1;
+
+		error = consumeWChars(lexer);
+		if (error != CSS_OK)
+			return error;
+
+		/* Fall through */
+	case Quote:
+		lexer->substate = Quote;
+
+		cptr = parserutils_inputstream_peek(lexer->input,
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF) {
+			/* Rewind to "url(" */
+			lexer->bytesReadForToken = lexer->context.bytesForURL;
+			lexer->token.data.len = lexer->context.dataLenForURL;
+			return emitToken(lexer, CSS_TOKEN_FUNCTION, token);
+		}
+
+		c = *((uint8_t *) cptr);
+
+		if (c == '"' || c == '\'') {
+			APPEND(lexer, cptr, clen);
+
+			lexer->context.first = c;
+
+			goto string;
+		}
+
+		/* Potential minor optimisation: If string is more common, 
+		 * then fall through to that state and branch for the URL 
+		 * state. Need to investigate a reasonably large corpus of 
+		 * real-world data to determine if this is worthwhile. */
+
+		/* Fall through */
+	case URL:
+		lexer->substate = URL;
+
+		error = consumeURLChars(lexer);
+		if (error != CSS_OK)
+			return error;
+
+		lexer->context.lastWasCR = false;
+
+		/* Fall through */
+	case W2:
+	w2:
+		lexer->substate = W2;
+
+		error = consumeWChars(lexer);
+		if (error != CSS_OK)
+			return error;
+
+		/* Fall through */
+	case RParen:
+		lexer->substate = RParen;
+
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF) {
+			/* Rewind to "url(" */
+			lexer->bytesReadForToken = lexer->context.bytesForURL;
+			lexer->token.data.len = lexer->context.dataLenForURL;
+			return emitToken(lexer, CSS_TOKEN_FUNCTION, token);
+		}
+
+		c = *((uint8_t *) cptr);
+
+		if (c != ')') {
+			/* Rewind to "url(" */
+			lexer->bytesReadForToken = lexer->context.bytesForURL;
+			lexer->token.data.len = lexer->context.dataLenForURL;
+			return emitToken(lexer, CSS_TOKEN_FUNCTION, token);
+		}
+
+		APPEND(lexer, cptr, clen);
+		break;
+	case String:
+	string:
+		lexer->substate = String;
+
+		error = consumeString(lexer);
+		if (error != CSS_OK && error != CSS_EOF)
+			return error;
+
+		/* EOF gets handled in RParen */
+
+		lexer->context.lastWasCR = false;
+
+		goto w2;
+	}
+
+	return emitToken(lexer, CSS_TOKEN_URI, token);
+}
+
+css_error UnicodeRange(css_lexer *lexer, const css_token **token)
+{
+	css_token *t = &lexer->token;
+	uintptr_t cptr = PARSERUTILS_INPUTSTREAM_OOD; /* GCC: shush */
+	uint8_t c = 0; /* GCC: shush */
+	size_t clen;
+	enum { Initial = 0, MoreDigits = 1 };
+
+	/* UNICODE-RANGE = [Uu] '+' [0-9a-fA-F?]{1,6}(-[0-9a-fA-F]{1,6})?
+	 * 
+	 * "U+" has been consumed.
+	 */
+
+	switch (lexer->substate) {
+	case Initial:
+		/* Attempt to consume 6 hex digits (or question marks) */
+		for (; lexer->context.hexCount < 6; lexer->context.hexCount++) {
+			cptr = parserutils_inputstream_peek(lexer->input,
+					lexer->bytesReadForToken, &clen);
+			if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+				return CSS_NEEDDATA;
+
+			if (cptr == PARSERUTILS_INPUTSTREAM_EOF) {
+				if (lexer->context.hexCount == 0) {
+					/* Remove '+' */
+					lexer->bytesReadForToken -= 1;
+					t->data.len -= 1;
+					/* u == IDENT */
+					return emitToken(lexer, 
+							CSS_TOKEN_IDENT, token);
+				} else {
+					return emitToken(lexer, 
+						CSS_TOKEN_UNICODE_RANGE, token);
+				}
+			}
+
+			c = *((uint8_t *) cptr);
+
+			if (isHex(c) || c == '?') {
+				APPEND(lexer, cptr, clen);
+			} else {
+				break;
+			}
+		}
+
+		if (lexer->context.hexCount == 0) {
+			/* We didn't consume any valid Unicode Range digits */
+			/* Remove the '+' */
+			lexer->bytesReadForToken -= 1;
+			t->data.len -= 1;
+			/* 'u' == IDENT */
+			return emitToken(lexer, CSS_TOKEN_IDENT, token);
+		} 
+
+		if (lexer->context.hexCount == 6) {
+			/* Consumed 6 valid characters. Look for '-' */
+			cptr = parserutils_inputstream_peek(lexer->input, 
+					lexer->bytesReadForToken, &clen);
+			if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+				return CSS_NEEDDATA;
+
+			if (cptr == PARSERUTILS_INPUTSTREAM_EOF)
+				return emitToken(lexer, 
+						CSS_TOKEN_UNICODE_RANGE, token);
+
+			c = *((uint8_t *) cptr);
+		}
+
+		/* If we've got a '-', then we may have a 
+		 * second range component */
+		if (c != '-') {
+			/* Reached the end of the range */
+			return emitToken(lexer, CSS_TOKEN_UNICODE_RANGE, token);
+		}
+
+		APPEND(lexer, cptr, clen);
+
+		/* Reset count for next set of digits */
+		lexer->context.hexCount = 0;
+
+		/* Fall through */
+	case MoreDigits:
+		lexer->substate = MoreDigits;
+
+		/* Consume up to 6 hex digits */
+		for (; lexer->context.hexCount < 6; lexer->context.hexCount++) {
+			cptr = parserutils_inputstream_peek(lexer->input, 
+					lexer->bytesReadForToken, &clen);
+			if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+				return CSS_NEEDDATA;
+
+			if (cptr == PARSERUTILS_INPUTSTREAM_EOF) {
+				if (lexer->context.hexCount == 0) {
+					/* Remove '-' */
+					lexer->bytesReadForToken -= 1;
+					t->data.len -= 1;
+				}
+
+				return emitToken(lexer, 
+						CSS_TOKEN_UNICODE_RANGE, token);
+			}
+
+			c = *((uint8_t *) cptr);
+
+			if (isHex(c)) {
+				APPEND(lexer, cptr, clen);
+			} else {
+				break;
+			}
+		}
+
+		if (lexer->context.hexCount == 0) {
+			/* No hex digits consumed. Remove '-' */
+			lexer->bytesReadForToken -= 1;
+			t->data.len -= 1;
+		}
+	}
+
+	return emitToken(lexer, CSS_TOKEN_UNICODE_RANGE, token);
+}
+
+/******************************************************************************
+ * Input consumers                                                            *
+ ******************************************************************************/
+
+css_error consumeDigits(css_lexer *lexer)
+{
+	uintptr_t cptr;
+	uint8_t c;
+	size_t clen;
+
+	/* digit = [0-9] */
+
+	/* Consume all digits */
+	do {
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF)
+			return CSS_OK;
+
+		c = *((uint8_t *) cptr);
+
+		if (isDigit(c)) {
+			APPEND(lexer, cptr, clen);
+		}
+	} while (isDigit(c));
+
+	return CSS_OK;
+}
+
+css_error consumeEscape(css_lexer *lexer, bool nl)
+{
+	uintptr_t cptr, sptr;
+	uint8_t c;
+	size_t clen, slen;
+	css_error error;
+
+	/* escape = unicode | '\' [^\n\r\f0-9a-fA-F] 
+	 * 
+	 * The '\' has been consumed.
+	 */
+
+	cptr = parserutils_inputstream_peek(lexer->input, 
+			lexer->bytesReadForToken, &clen);
+	if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+		return CSS_NEEDDATA;
+
+	if (cptr == PARSERUTILS_INPUTSTREAM_EOF)
+		return CSS_EOF;
+
+	c = *((uint8_t *) cptr);
+
+	if (!nl && (c == '\n' || c == '\r' || c == '\f')) {
+		/* These are not permitted */
+		return CSS_INVALID;
+	}
+
+	/* Create unescaped buffer, if it doesn't already exist */
+	if (lexer->unescapedTokenData == NULL) {
+		lexer->unescapedTokenData = 
+			parserutils_buffer_create(lexer->alloc, lexer->pw);
+		if (lexer->unescapedTokenData == NULL)
+			return CSS_NOMEM;
+	}
+
+	/* If this is the first escaped character we've seen for this token,
+	 * we must copy the characters we've read to the unescaped buffer */
+	if (!lexer->escapeSeen) {
+		if (lexer->bytesReadForToken > 1) {
+			sptr = parserutils_inputstream_peek(
+					lexer->input, 0, &slen);
+
+			assert(sptr != PARSERUTILS_INPUTSTREAM_EOF && 
+					sptr != PARSERUTILS_INPUTSTREAM_OOD);
+
+			/* -1 to skip '\\' */
+			error = css_error_from_parserutils_error(
+				parserutils_buffer_append(
+					lexer->unescapedTokenData, 
+					(const uint8_t *) sptr, 
+					lexer->bytesReadForToken - 1));
+			if (error != CSS_OK)
+				return error;
+		}
+
+		lexer->token.data.len = lexer->bytesReadForToken - 1;
+		lexer->escapeSeen = true;
+	}
+
+	if (isHex(c)) {
+		lexer->bytesReadForToken += clen;
+
+		error = consumeUnicode(lexer, charToHex(c));
+		if (error != CSS_OK) {
+			/* Rewind for next time */
+			lexer->bytesReadForToken -= clen;
+		}
+
+		return error;
+	}
+
+	/* If we're handling escaped newlines, convert CR(LF)? to LF */
+	if (nl && c == '\r') {
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken + clen, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF) {
+			c = '\n';
+			APPEND(lexer, &c, 1);
+
+			lexer->currentCol = 1;
+			lexer->currentLine++;
+
+			return CSS_OK;
+		}
+
+		c = *((uint8_t *) cptr);
+
+		if (c == '\n') {
+			APPEND(lexer, &c, 1);
+			/* And skip the '\r' in the input */
+			lexer->bytesReadForToken += clen;
+
+			lexer->currentCol = 1;
+			lexer->currentLine++;
+
+			return CSS_OK;
+		}
+	} else if (nl && (c == '\n' || c == '\f')) {
+		/* APPEND will increment this appropriately */
+		lexer->currentCol = 0;
+		lexer->currentLine++;
+	} else if (c != '\n' && c != '\r' && c != '\f') {
+		lexer->currentCol++;
+	}
+
+	/* Append the unescaped character */
+	APPEND(lexer, cptr, clen);
+
+	return CSS_OK;
+}
+
+css_error consumeNMChars(css_lexer *lexer)
+{
+	uintptr_t cptr;
+	uint8_t c;
+	size_t clen;
+	css_error error;
+
+	/* nmchar = [a-zA-Z] | '-' | '_' | nonascii | escape */
+
+	do {
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF)
+			return CSS_OK;
+
+		c = *((uint8_t *) cptr);
+
+		if (startNMChar(c) && c != '\\') {
+			APPEND(lexer, cptr, clen);
+		}
+
+		if (c == '\\') {
+			lexer->bytesReadForToken += clen;
+
+			error = consumeEscape(lexer, false);
+			if (error != CSS_OK) {
+				/* Rewind '\\', so we do the 
+				 * right thing next time */
+				lexer->bytesReadForToken -= clen;
+
+				/* Convert either EOF or INVALID into OK.
+				 * This will cause the caller to believe that
+				 * all NMChars in the sequence have been 
+				 * processed (and thus proceed to the next
+				 * state). Eventually, the '\\' will be output
+				 * as a CHAR. */
+				if (error == CSS_EOF || error == CSS_INVALID)
+					return CSS_OK;
+
+				return error;
+			}
+		}
+	} while (startNMChar(c));
+
+	return CSS_OK;
+}
+
+css_error consumeString(css_lexer *lexer)
+{
+	uintptr_t cptr;
+	uint8_t c;
+	size_t clen;
+	uint8_t quote = lexer->context.first;
+	uint8_t permittedquote = (quote == '"') ? '\'' : '"';
+	css_error error;
+
+	/* string = '"' (stringchar | "'")* '"' | "'" (stringchar | '"')* "'"
+	 *
+	 * The open quote has been consumed.
+	 */
+
+	do {
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF)
+			return CSS_EOF;
+
+		c = *((uint8_t *) cptr);
+
+		if (c == permittedquote) {
+			APPEND(lexer, cptr, clen);
+		} else if (startStringChar(c)) {
+			error = consumeStringChars(lexer);
+			if (error != CSS_OK)
+				return error;
+		} else if (c != quote) {
+			/* Invalid character in string -- skip */
+			lexer->bytesReadForToken += clen;
+		}
+	} while(c != quote);
+
+	/* Append closing quote to token data */
+	APPEND(lexer, cptr, clen);
+
+	return CSS_OK;
+}
+
+css_error consumeStringChars(css_lexer *lexer)
+{
+	uintptr_t cptr;
+	uint8_t c;
+	size_t clen;
+	css_error error;
+
+	/* stringchar = urlchar | ' ' | ')' | '\' nl */
+
+	do {
+		cptr = parserutils_inputstream_peek(lexer->input,
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF)
+			return CSS_OK;
+
+		c = *((uint8_t *) cptr);
+
+		if (startStringChar(c) && c != '\\') {
+			APPEND(lexer, cptr, clen);
+		}
+
+		if (c == '\\') {
+			lexer->bytesReadForToken += clen;
+
+			error = consumeEscape(lexer, true);
+			if (error != CSS_OK) {
+				/* Rewind '\\', so we do the 
+				 * right thing next time. */
+				lexer->bytesReadForToken -= clen;
+
+				/* Convert EOF to OK. This causes the caller
+				 * to believe that all StringChars have been
+				 * processed. Eventually, the '\\' will be
+				 * output as a CHAR. */
+				if (error == CSS_EOF)
+					return CSS_OK;
+
+				return error;
+			}
+		}
+	} while (startStringChar(c));
+
+	return CSS_OK;
+
+}
+
+css_error consumeUnicode(css_lexer *lexer, uint32_t ucs)
+{
+	uintptr_t cptr = PARSERUTILS_INPUTSTREAM_OOD; /* GCC: shush */
+	uint8_t c = 0;
+	size_t clen;
+	uint8_t utf8[6];
+	uint8_t *utf8ptr = utf8;
+	size_t utf8len = sizeof(utf8);
+	size_t bytesReadInit = lexer->bytesReadForToken;
+	int count;
+	parserutils_error error;
+
+	/* unicode = '\' [0-9a-fA-F]{1,6} wc? 
+	 *
+	 * The '\' and the first digit have been consumed.
+	 */
+
+	/* Attempt to consume a further five hex digits */
+	for (count = 0; count < 5; count++) {
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD) {
+			/* Rewind what we've read */
+			lexer->bytesReadForToken = bytesReadInit;
+			return CSS_NEEDDATA;
+		}
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF)
+			break;
+
+		c = *((uint8_t *) cptr);
+
+		if (isHex(c)) {
+			lexer->bytesReadForToken += clen;
+
+			ucs = (ucs << 4) | charToHex(c);
+		} else {
+			break;
+		}
+	}
+
+	/* Convert our UCS4 character to UTF-8 */
+	error = parserutils_charset_utf8_from_ucs4(ucs, &utf8ptr, &utf8len);
+	assert(error == PARSERUTILS_OK);
+
+	/* Append it to the token data (unescaped buffer already set up) */
+	/* We can't use the APPEND() macro here as we want to rewind correctly
+	 * on error. Additionally, lexer->bytesReadForToken has already been
+	 * advanced */
+	error = appendToTokenData(lexer, (const uint8_t *) utf8, 
+			sizeof(utf8) - utf8len);
+	if (error != CSS_OK) {
+		/* Rewind what we've read */
+		lexer->bytesReadForToken = bytesReadInit;
+		return error;
+	}
+
+	/* Finally, attempt to skip a whitespace character */
+	if (cptr == PARSERUTILS_INPUTSTREAM_EOF)
+		return CSS_OK;
+
+	if (isSpace(c)) {
+		lexer->bytesReadForToken += clen;
+	}
+
+	/* +2 for '\' and first digit */
+	lexer->currentCol += lexer->bytesReadForToken - bytesReadInit + 2;
+
+	return CSS_OK;
+}
+
+css_error consumeURLChars(css_lexer *lexer)
+{
+	uintptr_t cptr;
+	uint8_t c;
+	size_t clen;
+	css_error error;
+
+	/* urlchar = [\t!#-&(*-~] | nonascii | escape */
+
+	do {
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF)
+			return CSS_OK;
+
+		c = *((uint8_t *) cptr);
+
+		if (startURLChar(c) && c != '\\') {
+			APPEND(lexer, cptr, clen);
+		}
+
+		if (c == '\\') {
+			lexer->bytesReadForToken += clen;
+
+			error = consumeEscape(lexer, false);
+			if (error != CSS_OK) {
+				/* Rewind '\\', so we do the
+				 * right thing next time */
+				lexer->bytesReadForToken -= clen;
+
+				/* Convert either EOF or INVALID into OK.
+				 * This will cause the caller to believe that
+				 * all URLChars in the sequence have been 
+				 * processed (and thus proceed to the next
+				 * state). Eventually, the '\\' will be output
+				 * as a CHAR. */
+				if (error == CSS_EOF || error == CSS_INVALID)
+					return CSS_OK;
+
+				return error;
+			}
+		}
+	} while (startURLChar(c));
+
+	return CSS_OK;
+}
+
+css_error consumeWChars(css_lexer *lexer)
+{
+	uintptr_t cptr;
+	uint8_t c;
+	size_t clen;
+
+	do {
+		cptr = parserutils_inputstream_peek(lexer->input, 
+				lexer->bytesReadForToken, &clen);
+		if (cptr == PARSERUTILS_INPUTSTREAM_OOD)
+			return CSS_NEEDDATA;
+
+		if (cptr == PARSERUTILS_INPUTSTREAM_EOF)
+			return CSS_OK;
+
+		c = *((uint8_t *) cptr);
+
+		if (isSpace(c)) {
+			APPEND(lexer, cptr, clen);
+		}
+
+		if (c == '\n' || c == '\f') {
+			lexer->currentCol = 1;
+			lexer->currentLine++;
+		}
+
+		if (lexer->context.lastWasCR && c != '\n') {
+			lexer->currentCol = 1;
+			lexer->currentLine++;
+		}
+		lexer->context.lastWasCR = (c == '\r');
+	} while (isSpace(c));
+
+	if (lexer->context.lastWasCR) {
+		lexer->currentCol = 1;
+		lexer->currentLine++;
+	}
+
+	return CSS_OK;
+}
+
+/******************************************************************************
+ * More utility routines                                                      *
+ ******************************************************************************/
+
+uint32_t charToHex(uint8_t c)
+{
+	switch (c) {
+	case 'a': case 'A':
+		return 0xa;
+	case 'b': case 'B':
+		return 0xb;
+	case 'c': case 'C':
+		return 0xc;
+	case 'd': case 'D':
+		return 0xd;
+	case 'e': case 'E':
+		return 0xe;
+	case 'f': case 'F':
+		return 0xf;
+	case '0':
+		return 0x0;
+	case '1':
+		return 0x1;
+	case '2':
+		return 0x2;
+	case '3':
+		return 0x3;
+	case '4':
+		return 0x4;
+	case '5':
+		return 0x5;
+	case '6':
+		return 0x6;
+	case '7':
+		return 0x7;
+	case '8':
+		return 0x8;
+	case '9':
+		return 0x9;
+	}
+
+	return 0;
+}
+
+bool startNMChar(uint8_t c)
+{
+	return c == '_' || ('a' <= c && c <= 'z') || ('A' <= c && c <= 'Z') || 
+		('0' <= c && c <= '9') || c == '-' || c >= 0x80 || c == '\\';
+}
+
+bool startNMStart(uint8_t c)
+{
+	return c == '_' || ('a' <= c && c <= 'z') || ('A' <= c && c <= 'Z') ||
+		c >= 0x80 || c == '\\';
+}
+
+bool startStringChar(uint8_t c)
+{
+	return startURLChar(c) || c == ' ' || c == ')';
+}
+
+bool startURLChar(uint8_t c)
+{
+	return c == '\t' || c == '!' || ('#' <= c && c <= '&') || c == '(' ||
+		('*' <= c && c <= '~') || c >= 0x80 || c == '\\';
+}
+
+bool isDigit(uint8_t c)
+{
+	return '0' <= c && c <= '9';
+}
+
+bool isHex(uint8_t c)
+{
+	return isDigit(c) || ('a' <= c && c <= 'f') || ('A' <= c && c <= 'F');
+}
+
+bool isSpace(uint8_t c)
+{
+	return c == ' ' || c == '\r' || c == '\n' || c == '\f' || c == '\t';
+}
+
diff --git a/src/lex/lex.h b/src/lex/lex.h
new file mode 100644
index 0000000..150823e
--- /dev/null
+++ b/src/lex/lex.h
@@ -0,0 +1,67 @@
+/*
+ * This file is part of LibCSS.
+ * Licensed under the MIT License,
+ *                http://www.opensource.org/licenses/mit-license.php
+ * Copyright 2008 John-Mark Bell <jmb@netsurf-browser.org>
+ */
+
+#ifndef css_lex_lex_h_
+#define css_lex_lex_h_
+
+#include <libcss/functypes.h>
+#include <libcss/types.h>
+
+#include <parserutils/input/inputstream.h>
+
+typedef struct css_lexer css_lexer;
+
+/**
+ * Lexer option types
+ */
+typedef enum css_lexer_opttype {
+	CSS_LEXER_EMIT_COMMENTS,
+} css_lexer_opttype;
+
+/**
+ * Lexer option parameters
+ */
+typedef union css_lexer_optparams {
+	bool emit_comments;
+} css_lexer_optparams;
+
+/**
+ * Token type
+ */
+typedef enum css_token_type{ 
+	CSS_TOKEN_IDENT, CSS_TOKEN_ATKEYWORD, CSS_TOKEN_STRING,
+	CSS_TOKEN_HASH, CSS_TOKEN_NUMBER, CSS_TOKEN_PERCENTAGE, 
+	CSS_TOKEN_DIMENSION, CSS_TOKEN_URI, CSS_TOKEN_UNICODE_RANGE, 
+	CSS_TOKEN_CDO, CSS_TOKEN_CDC, CSS_TOKEN_S, CSS_TOKEN_COMMENT,
+	CSS_TOKEN_FUNCTION, CSS_TOKEN_INCLUDES, CSS_TOKEN_DASHMATCH,
+	CSS_TOKEN_PREFIXMATCH, CSS_TOKEN_SUFFIXMATCH, CSS_TOKEN_SUBSTRINGMATCH,
+	CSS_TOKEN_CHAR, CSS_TOKEN_EOF 
+} css_token_type;
+
+/**
+ * Token object
+ */
+typedef struct css_token {
+	css_token_type type;
+
+	css_string data;
+
+	uint32_t col;
+	uint32_t line;
+} css_token;
+
+css_lexer *css_lexer_create(parserutils_inputstream *input, 
+		css_alloc alloc, void *pw);
+void css_lexer_destroy(css_lexer *lexer);
+
+css_error css_lexer_setopt(css_lexer *lexer, css_lexer_opttype type, 
+		css_lexer_optparams *params);
+
+css_error css_lexer_get_token(css_lexer *lexer, const css_token **token);
+
+#endif
+
diff --git a/src/libcss.c b/src/libcss.c
new file mode 100644
index 0000000..b5b99c9
--- /dev/null
+++ b/src/libcss.c
@@ -0,0 +1,50 @@
+/*
+ * This file is part of LibCSS.
+ * Licensed under the MIT License,
+ *                http://www.opensource.org/licenses/mit-license.php
+ * Copyright 2007 John-Mark Bell <jmb@netsurf-browser.org>
+ */
+
+#include <parserutils/parserutils.h>
+
+#include <libcss/libcss.h>
+
+#include "utils/parserutilserror.h"
+
+/**
+ * Initialise the CSS library for use.
+ *
+ * This _must_ be called before using any LibCSS functions
+ *
+ * \param aliases_file  Pointer to name of file containing encoding alias data
+ * \param alloc         Pointer to (de)allocation function
+ * \param pw            Pointer to client-specific private data (may be NULL)
+ * \return CSS_OK on success, applicable error otherwise.
+ */
+css_error css_initialise(const char *aliases_file,
+		css_alloc alloc, void *pw)
+{
+	if (aliases_file == NULL || alloc == NULL)
+		return CSS_BADPARM;
+
+	return css_error_from_parserutils_error(
+			parserutils_initialise(aliases_file, alloc, pw));
+}
+
+/**
+ * Clean up after LibCSS
+ *
+ * \param alloc  Pointer to (de)allocation function
+ * \param pw     Pointer to client-specific private data (may be NULL)
+ * \return CSS_OK on success, applicable error otherwise.
+ */
+css_error css_finalise(css_alloc alloc, void *pw)
+{
+	if (alloc == NULL)
+		return CSS_BADPARM;
+
+	return css_error_from_parserutils_error(
+			parserutils_finalise(alloc, pw));
+}
+
+
diff --git a/src/utils/Makefile b/src/utils/Makefile
new file mode 100644
index 0000000..912590c
--- /dev/null
+++ b/src/utils/Makefile
@@ -0,0 +1,46 @@
+# Child makefile fragment
+#
+# Toolchain is provided by top-level makefile
+#
+# Variables provided by top-level makefile
+#
+# COMPONENT		The name of the component
+# EXPORT		The location of the export directory
+# TOP			The location of the source tree root
+# RELEASEDIR		The place to put release objects
+# DEBUGDIR		The place to put debug objects
+#
+# do_include		Canned command sequence to include a child makefile
+#
+# Variables provided by parent makefile:
+#
+# DIR			The name of the directory we're in, relative to $(TOP)
+#
+# Variables we can manipulate:
+#
+# ITEMS_CLEAN		The list of items to remove for "make clean"
+# ITEMS_DISTCLEAN	The list of items to remove for "make distclean"
+# TARGET_TESTS		The list of target names to run for "make test"
+#
+# SOURCES		The list of sources to build for $(COMPONENT)
+#
+# Plus anything from the toolchain
+
+# Push parent directory onto the directory stack
+sp             := $(sp).x
+dirstack_$(sp) := $(d)
+d              := $(DIR)
+
+# Sources
+SRCS_$(d) := errors.c
+
+# Append to sources for component
+SOURCES += $(addprefix $(d), $(SRCS_$(d)))
+
+# Now include any children we may have
+MAKE_INCLUDES := $(wildcard $(d)*/Makefile)
+$(eval $(foreach INC, $(MAKE_INCLUDES), $(call do_include,$(INC))))
+
+# Finally, pop off the directory stack
+d  := $(dirstack_$(sp))
+sp := $(basename $(sp))
diff --git a/src/utils/errors.c b/src/utils/errors.c
new file mode 100644
index 0000000..ac7dd42
--- /dev/null
+++ b/src/utils/errors.c
@@ -0,0 +1,80 @@
+/*
+ * This file is part of LibCSS.
+ * Licensed under the MIT License,
+ *                http://www.opensource.org/licenses/mit-license.php
+ * Copyright 2007 John-Mark Bell <jmb@netsurf-browser.org>
+ */
+
+#include <string.h>
+
+#include <libcss/errors.h>
+
+/**
+ * Convert a LibCSS error code to a string
+ *
+ * \param error  The error code to convert
+ * \return Pointer to string representation of error, or NULL if unknown.
+ */
+const char *css_error_to_string(css_error error)
+{
+	const char *result = NULL;
+
+	switch (error) {
+	case CSS_OK:
+		result = "No error";
+		break;
+	case CSS_NOMEM:
+		result = "Insufficient memory";
+		break;
+	case CSS_BADPARM:
+		result = "Bad parameter";
+		break;
+	case CSS_INVALID:
+		result = "Invalid input";
+		break;
+	case CSS_FILENOTFOUND:
+		result = "File not found";
+		break;
+	case CSS_NEEDDATA:
+		result = "Insufficient data";
+		break;
+	case CSS_BADCHARSET:
+		result = "BOM and @charset mismatch";
+		break;
+	case CSS_EOF:
+		result = "EOF encountered";
+		break;
+	}
+
+	return result;
+}
+
+/**
+ * Convert a string representation of an error name to a LibCSS error code
+ *
+ * \param str  String containing error name
+ * \param len  Length of string (bytes)
+ * \return LibCSS error code, or CSS_OK if unknown
+ */
+css_error css_error_from_string(const char *str, size_t len)
+{
+	if (strncmp(str, "CSS_OK", len) == 0) {
+		return CSS_OK;
+	} else if (strncmp(str, "CSS_NOMEM", len) == 0) {
+		return CSS_NOMEM;
+	} else if (strncmp(str, "CSS_BADPARM", len) == 0) {
+		return CSS_BADPARM;
+	} else if (strncmp(str, "CSS_INVALID", len) == 0) {
+		return CSS_INVALID;
+	} else if (strncmp(str, "CSS_FILENOTFOUND", len) == 0) {
+		return CSS_FILENOTFOUND;
+	} else if (strncmp(str, "CSS_NEEDDATA", len) == 0) {
+		return CSS_NEEDDATA;
+	} else if (strncmp(str, "CSS_BADCHARSET", len) == 0) {
+		return CSS_BADCHARSET;
+	} else if (strncmp(str, "CSS_EOF", len) == 0) {
+		return CSS_EOF;
+	}
+
+	return CSS_OK;
+}
diff --git a/src/utils/parserutilserror.h b/src/utils/parserutilserror.h
new file mode 100644
index 0000000..d2cffb6
--- /dev/null
+++ b/src/utils/parserutilserror.h
@@ -0,0 +1,29 @@
+/*
+ * This file is part of LibCSS.
+ * Licensed under the MIT License,
+ *                http://www.opensource.org/licenses/mit-license.php
+ * Copyright 2007 John-Mark Bell <jmb@netsurf-browser.org>
+ */
+
+#ifndef css_utils_parserutilserror_h_
+#define css_utils_parserutilserror_h_
+
+#include <parserutils/errors.h>
+
+#include <libcss/errors.h>
+
+/**
+ * Convert a ParserUtils error into a LibCSS error
+ * 
+ * \param error  The ParserUtils error to convert
+ * \return The corresponding LibCSS error
+ */
+static inline css_error css_error_from_parserutils_error(
+		parserutils_error error)
+{
+	/* Currently, there's a 1:1 mapping, so we've nothing to do */
+	return (css_error) error;
+}
+
+#endif
+
diff --git a/src/utils/utils.h b/src/utils/utils.h
new file mode 100644
index 0000000..ac19e59
--- /dev/null
+++ b/src/utils/utils.h
@@ -0,0 +1,28 @@
+/*
+ * This file is part of LibCSS.
+ * Licensed under the MIT License,
+ *                http://www.opensource.org/licenses/mit-license.php
+ * Copyright 2007 John-Mark Bell <jmb@netsurf-browser.org>
+ */
+
+#ifndef css_utils_h_
+#define css_utils_h_
+
+#ifndef max
+#define max(a,b) ((a)>(b)?(a):(b))
+#endif
+
+#ifndef min
+#define min(a,b) ((a)<(b)?(a):(b))
+#endif
+
+#ifndef SLEN
+/* Calculate length of a string constant */
+#define SLEN(s) (sizeof((s)) - 1) /* -1 for '\0' */
+#endif
+
+#ifndef UNUSED
+#define UNUSED(x) ((x)=(x))
+#endif
+
+#endif