Functions for parsing rawml markup. More...

#include <stdlib.h>
#include <string.h>
#include <ctype.h>
#include "parse_rawml.h"
#include "util.h"
#include "opf.h"
#include "structure.h"
#include "index.h"
#include "debug.h"

Macros
#define	_GNU_SOURCE 1

#define	__USE_BSD /* for strdup on linux/glibc */

Functions
size_t	mobi_get_rawlink_location (const MOBIRawml *rawml, const uint32_t pos_fid, const uint32_t pos_off)
	Convert kindle:pos:fid:x:off:y to offset in rawml raw text file. More...

MOBI_RET	mobi_search_links_kf7 (MOBIResult result, const unsigned char data_start, const unsigned char *data_end)
	Find first occurence of attribute to be replaced in KF7 html. More...

MOBI_RET	mobi_find_attrvalue (MOBIResult result, const unsigned char data_start, const unsigned char data_end, const MOBIFiletype type, const char needle)
	Find first occurence of markup attribute with given value. More...

MOBI_RET	mobi_find_attrname (MOBIResult result, const unsigned char data_start, const unsigned char data_end, const char attrname)
	Find first occurence of markup attribute with given name. More...

MOBI_RET	mobi_search_links_kf8 (MOBIResult result, const unsigned char data_start, const unsigned char *data_end, const MOBIFiletype type)
	Find first occurence of attribute part to be replaced in KF8 html/css. More...

size_t	mobi_get_attribute_value (char value, const unsigned char data, const size_t size, const char *attribute, bool only_quoted)
	Get value and offset of the first found attribute with given name. More...

size_t	mobi_get_aid_offset (const MOBIPart html, const char aid)
	Get offset of the given value of an "aid" attribute in a given part. More...

MOBI_RET	mobi_get_offset_by_posoff (uint32_t file_number, size_t offset, const MOBIRawml *rawml, const size_t pos_fid, const size_t pos_off)
	Convert kindle:pos:fid:x:off:y to skeleton part number and offset from the beginning of the part. More...

MOBI_RET	mobi_get_aid_by_offset (char aid, const MOBIPart html, const size_t offset)
	Get value of the closest "aid" attribute following given offset in a given part. More...

MOBI_RET	mobi_get_id_by_offset (char id, const MOBIPart html, const size_t offset, MOBIAttrType *pref_attr)
	Get value of the closest "id" or "name" attribute following given offset in a given part. More...

MOBI_RET	mobi_get_aid_by_posoff (uint32_t file_number, char aid, const MOBIRawml *rawml, const size_t pos_fid, const size_t pos_off)
	Convert kindle:pos:fid:x:off:y to html file number and closest "aid" attribute following the position. More...

MOBI_RET	mobi_get_id_by_posoff (uint32_t file_number, char id, const MOBIRawml rawml, const size_t pos_fid, const size_t pos_off, MOBIAttrType pref_attr)
	Convert kindle:pos:fid:x:off:y to html file number and closest "id" attribute following the position. More...

MOBI_RET	mobi_reconstruct_resources (const MOBIData m, MOBIRawml rawml)
	Parse resource records (images, fonts etc), determine their type, link to rawml. More...

MOBI_RET	mobi_process_replica (unsigned char pdf, const char text, size_t *length)
	Parse Replica Print ebook (azw4). Extract pdf. More...

MOBI_RET	mobi_reconstruct_flow (MOBIRawml rawml, const char text, const size_t length)
	Parse raw text into flow parts. More...

MOBI_RET	mobi_reconstruct_parts (MOBIRawml *rawml)
	Parse raw html into html parts. Use index entries if present to parse file. More...

MOBI_RET	mobi_get_filepos_array (MOBIArray links, const MOBIPart part)
	Scan html part and build array of filepos link target offsets. More...

MOBI_RET	mobi_get_ncx_filepos_array (MOBIArray links, const MOBIRawml rawml)
	Scan ncx part and build array of filepos link target offsets. More...

MOBI_RET	mobi_posfid_to_link (char link, const MOBIRawml rawml, const char value, MOBIAttrType pref_attr)
	Replace kindle:pos link with html href. More...

MOBI_RET	mobi_flow_to_link (char link, const MOBIRawml rawml, const char *value)
	Replace kindle:flow link with html href. More...

MOBI_RET	mobi_embed_to_link (char link, const MOBIRawml rawml, const char *value)
	Replace kindle:embed link with html href. More...

MOBI_RET	mobi_reconstruct_links_kf8 (const MOBIRawml *rawml)
	Replace offset-links with html-links in KF8 markup. More...

MOBI_RET	mobi_reconstruct_infl (char outstring, const MOBIIndx infl, const MOBIIndexEntry *orth_entry)
	Get infl index markup for given orth entry. More...

MOBI_RET	mobi_reconstruct_infl_v1 (char outstring, MOBITrie const infl_tree, const MOBIIndexEntry *orth_entry)
	Get infl index markup for given orth entry. More...

MOBI_RET	mobi_reconstruct_orth (const MOBIRawml rawml, MOBIFragment first, size_t *new_size)
	Insert orth index markup to linked list of fragments. More...

MOBI_RET	mobi_reconstruct_links_kf7 (const MOBIRawml *rawml)
	Replace offset-links with html-links in KF7 markup. Also reconstruct dictionary markup if present. More...

MOBI_RET	mobi_reconstruct_links (const MOBIRawml *rawml)
	Replace offset-links with html-links. More...

MOBI_RET	mobi_iterate_txtparts (MOBIRawml rawml, MOBI_RET(cb)(MOBIPart *))
	Call callback function for each text record. More...

MOBI_RET	mobi_markup_to_utf8 (MOBIPart *part)
	Convert MOBIPart part data to utf8. More...

MOBI_RET	mobi_strip_mobitags (MOBIPart *part)
	Strip unneeded tags from html. Currently only <aid> More...

MOBI_RET	mobi_parse_rawml (MOBIRawml rawml, const MOBIData m)
	Parse raw records into html flow parts, markup parts, resources and indices. More...

MOBI_RET	mobi_parse_rawml_opt (MOBIRawml rawml, const MOBIData m, bool parse_toc, bool parse_dict, bool reconstruct)
	Parse raw records into html flow parts, markup parts, resources and indices. Individual stages of the parsing may be turned on/off. More...

Detailed Description

Functions for parsing rawml markup.

This file is part of libmobi. Licensed under LGPL, either version 3, or any later. See http://www.gnu.org/licenses/

Function Documentation

◆ mobi_embed_to_link()

MOBI_RET mobi_embed_to_link	(	char *	link,
		const MOBIRawml *	rawml,
		const char *	value
	)

Replace kindle:embed link with html href.

Parameters

[in,out]	link	Memory area which will be filled with "resource00000.ext", including quotation marks
[in]	rawml	Structure rawml
[in]	value	String kindle:embed:0000?mime=type, with optional quotation marks

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_find_attrname()

MOBI_RET mobi_find_attrname	(	MOBIResult *	result,
		const unsigned char *	data_start,
		const unsigned char *	data_end,
		const char *	attrname
	)

Find first occurence of markup attribute with given name.

Parameters

[in,out]	result	MOBIResult structure will be filled with found data
[in]	data_start	Beginning of the memory area to search in
[in]	data_end	End of the memory area to search in
[in]	attrname	String to find (len < MOBI_ATTRNAME_MAXSIZE)

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_find_attrvalue()

MOBI_RET mobi_find_attrvalue	(	MOBIResult *	result,
		const unsigned char *	data_start,
		const unsigned char *	data_end,
		const MOBIFiletype	type,
		const char *	needle
	)

Find first occurence of markup attribute with given value.

Parameters

[in,out]	result	MOBIResult structure will be filled with found data
[in]	data_start	Beginning of the memory area to search in
[in]	data_end	End of the memory area to search in
[in]	type	Type of data (T_HTML or T_CSS)
[in]	needle	String to find (len <= MOBI_ATTRNAME_MAXSIZE)

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_flow_to_link()

MOBI_RET mobi_flow_to_link	(	char *	link,
		const MOBIRawml *	rawml,
		const char *	value
	)

Replace kindle:flow link with html href.

Parameters

[in,out]	link	Memory area which will be filled with "part00000.ext", including quotation marks
[in]	rawml	Structure rawml
[in]	value	String kindle:flow:0000?mime=type, without quotation marks

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_get_aid_by_offset()

MOBI_RET mobi_get_aid_by_offset	(	char *	aid,
		const MOBIPart *	html,
		const size_t	offset
	)

Get value of the closest "aid" attribute following given offset in a given part.

Parameters

[in,out]	aid	String value of "aid" attribute
[in]	html	MOBIPart html part
[in]	offset	Offset from the beginning of the part data

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_get_aid_by_posoff()

MOBI_RET mobi_get_aid_by_posoff	(	uint32_t *	file_number,
		char *	aid,
		const MOBIRawml *	rawml,
		const size_t	pos_fid,
		const size_t	pos_off
	)

Convert kindle:pos:fid:x:off:y to html file number and closest "aid" attribute following the position.

Parameters

[in,out]	file_number	Will be set to file number value
[in,out]	aid	String value of "aid" attribute
[in]	rawml	MOBIRawml parsed records structure
[in]	pos_fid	X value of pos:fid:x
[in]	pos_off	Y value of off:y

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_get_aid_offset()

size_t mobi_get_aid_offset	(	const MOBIPart *	html,
		const char *	aid
	)

Get offset of the given value of an "aid" attribute in a given part.

Parameters

[in]	aid	String value of "aid" attribute
[in]	html	MOBIPart html part

Returns: Offset from the beginning of the html part data, SIZE_MAX on failure

◆ mobi_get_attribute_value()

size_t mobi_get_attribute_value	(	char *	value,
		const unsigned char *	data,
		const size_t	size,
		const char *	attribute,
		bool	only_quoted
	)

Get value and offset of the first found attribute with given name.

Parameters

[in,out]	value	String value of the attribute, will be filled by the function, zero length if not found
[in]	data	Data to search in
[in]	size	Data size
[in]	attribute	Attribute name
[in]	only_quoted	Require the value to be quoted if true, allow no quotes (eg. filepos=00001) if false

Returns: Offset from the beginning of the data, SIZE_MAX if not found

◆ mobi_get_filepos_array()

MOBI_RET mobi_get_filepos_array	(	MOBIArray *	links,
		const MOBIPart *	part
	)

Scan html part and build array of filepos link target offsets.

Parameters

[in,out]	links	MOBIArray structure for link target offsets array
[in]	part	MOBIPart html part structure

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_get_id_by_offset()

MOBI_RET mobi_get_id_by_offset	(	char *	id,
		const MOBIPart *	html,
		const size_t	offset,
		MOBIAttrType *	pref_attr
	)

Get value of the closest "id" or "name" attribute following given offset in a given part.

Parameters

[in,out]	id	String value of found attribute
[in]	html	MOBIPart html part
[in]	offset	Offset from the beginning of the part data
[in,out]	pref_attr	Preferred attribute to link to (id or name)

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_get_id_by_posoff()

MOBI_RET mobi_get_id_by_posoff	(	uint32_t *	file_number,
		char *	id,
		const MOBIRawml *	rawml,
		const size_t	pos_fid,
		const size_t	pos_off,
		MOBIAttrType *	pref_attr
	)

Convert kindle:pos:fid:x:off:y to html file number and closest "id" attribute following the position.

Parameters

[in,out]	file_number	Will be set to file number value
[in,out]	id	String value of "id" attribute
[in]	rawml	MOBIRawml parsed records structure
[in]	pos_fid	X value of pos:fid:x
[in]	pos_off	Y value of off:y
[in,out]	pref_attr	Attribute to link to

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_get_ncx_filepos_array()

MOBI_RET mobi_get_ncx_filepos_array	(	MOBIArray *	links,
		const MOBIRawml *	rawml
	)

Scan ncx part and build array of filepos link target offsets.

Parameters

[in,out]	links	MOBIArray structure for link target offsets array
[in]	rawml	MOBIRawml parsed records structure

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_get_offset_by_posoff()

MOBI_RET mobi_get_offset_by_posoff	(	uint32_t *	file_number,
		size_t *	offset,
		const MOBIRawml *	rawml,
		const size_t	pos_fid,
		const size_t	pos_off
	)

Convert kindle:pos:fid:x:off:y to skeleton part number and offset from the beginning of the part.

Parameters

[in,out]	file_number	Will be set to file number value
[in,out]	offset	Offset from the beginning of the skeleton part
[in]	rawml	MOBIRawml parsed records structure
[in]	pos_fid	X value of pos:fid:x
[in]	pos_off	X value of pos:off:x

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_get_rawlink_location()

size_t mobi_get_rawlink_location	(	const MOBIRawml *	rawml,
		const uint32_t	pos_fid,
		const uint32_t	pos_off
	)

Convert kindle:pos:fid:x:off:y to offset in rawml raw text file.

Parameters

[in]	rawml	MOBIRawml parsed records structure
[in]	pos_fid	X value of pos:fid:x
[in]	pos_off	Y value of off:y

Returns: Offset in rawml buffer on success, SIZE_MAX otherwise

◆ mobi_iterate_txtparts()

MOBI_RET mobi_iterate_txtparts	(	MOBIRawml *	rawml,
		MOBI_RET()(MOBIPart )	cb
	)

Call callback function for each text record.

Parameters

[in,out]	rawml	Structure rawml will be filled with reconstructed parts and resources
[in,out]	cb	Callback function

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_markup_to_utf8()

MOBI_RET mobi_markup_to_utf8 ( MOBIPart * part )

Convert MOBIPart part data to utf8.

Parameters

[in,out] part MOBIPart part

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_posfid_to_link()

MOBI_RET mobi_posfid_to_link	(	char *	link,
		const MOBIRawml *	rawml,
		const char *	value,
		MOBIAttrType *	pref_attr
	)

Replace kindle:pos link with html href.

Parameters

[in,out]	link	Memory area which will be filled with "part00000.html#customid", including quotation marks
[in]	rawml	Structure rawml
[in]	value	String kindle:pos:fid:0000:off:0000000000, without quotation marks
[in,out]	pref_attr	Preferred attribute to link to (id or name)

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_process_replica()

MOBI_RET mobi_process_replica	(	unsigned char *	pdf,
		const char *	text,
		size_t *	length
	)

Parse Replica Print ebook (azw4). Extract pdf.

Todo:: Parse remaining data from the file

Parameters

[in,out]	pdf	Memory area will be filled with extracted pdf data
[in]	text	Raw decompressed text to be parsed
[in,out]	length	Text length. Will be updated with pdf_length on return

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_reconstruct_flow()

MOBI_RET mobi_reconstruct_flow	(	MOBIRawml *	rawml,
		const char *	text,
		const size_t	length
	)

Parse raw text into flow parts.

Parameters

[in,out]	rawml	Structure rawml->flow will be filled with parsed flow text parts
[in]	text	Raw decompressed text to be parsed
[in]	length	Text length

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_reconstruct_infl()

MOBI_RET mobi_reconstruct_infl	(	char *	outstring,
		const MOBIIndx *	infl,
		const MOBIIndexEntry *	orth_entry
	)

Get infl index markup for given orth entry.

Parameters

[in,out]	outstring	Reconstructed tag <idx:infl>
[in]	infl	MOBIIndx structure with parsed infl index
[in]	orth_entry	Orth index entry

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_reconstruct_infl_v1()

MOBI_RET mobi_reconstruct_infl_v1	(	char *	outstring,
		MOBITrie *const	infl_tree,
		const MOBIIndexEntry *	orth_entry
	)

Get infl index markup for given orth entry.

This function is inflections scheme used in older mobipocket dictionaries

Parameters

[in,out]	outstring	Reconstructed tag <idx:infl>
[in]	infl_tree	MOBITrie structure with inflection rules
[in]	orth_entry	Orth index entry

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_reconstruct_links()

MOBI_RET mobi_reconstruct_links ( const MOBIRawml * rawml )

Replace offset-links with html-links.

Parameters

[in,out] rawml Structure rawml will be filled with reconstructed parts and resources

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_reconstruct_links_kf7()

MOBI_RET mobi_reconstruct_links_kf7 ( const MOBIRawml * rawml )

Replace offset-links with html-links in KF7 markup. Also reconstruct dictionary markup if present.

Parameters

[in,out] rawml Structure rawml will be filled with reconstructed parts and resources

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_reconstruct_links_kf8()

MOBI_RET mobi_reconstruct_links_kf8 ( const MOBIRawml * rawml )

Replace offset-links with html-links in KF8 markup.

Parameters

[in,out] rawml Structure rawml will be filled with reconstructed parts and resources

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_reconstruct_orth()

MOBI_RET mobi_reconstruct_orth	(	const MOBIRawml *	rawml,
		MOBIFragment *	first,
		size_t *	new_size
	)

Insert orth index markup to linked list of fragments.

Parameters

[in]	rawml	Structure rawml contains orth index data
[in,out]	first	First element of the linked list
[in,out]	new_size	Counter to be updated with inserted fragments size

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_reconstruct_parts()

MOBI_RET mobi_reconstruct_parts ( MOBIRawml * rawml )

Parse raw html into html parts. Use index entries if present to parse file.

Parameters

[in,out] rawml Structure rawml->markup will be filled with reconstructed html parts

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_reconstruct_resources()

MOBI_RET mobi_reconstruct_resources	(	const MOBIData *	m,
		MOBIRawml *	rawml
	)

Parse resource records (images, fonts etc), determine their type, link to rawml.

Parameters

[in]	m	MOBIData structure with loaded Record(s) 0 headers
[in,out]	rawml	Structure rawml->resources will be filled with parsed resources metadata and linked records data

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_search_links_kf7()

MOBI_RET mobi_search_links_kf7	(	MOBIResult *	result,
		const unsigned char *	data_start,
		const unsigned char *	data_end
	)

Find first occurence of attribute to be replaced in KF7 html.

It searches for filepos and recindex attributes

Parameters

[in,out]	result	MOBIResult structure will be filled with found data
[in]	data_start	Beginning of the memory area to search in
[in]	data_end	End of the memory area to search in

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_search_links_kf8()

MOBI_RET mobi_search_links_kf8	(	MOBIResult *	result,
		const unsigned char *	data_start,
		const unsigned char *	data_end,
		const MOBIFiletype	type
	)

Find first occurence of attribute part to be replaced in KF8 html/css.

It searches for "kindle:" value in attributes

Parameters

[in,out]	result	MOBIResult structure will be filled with found data
[in]	data_start	Beginning of the memory area to search in
[in]	data_end	End of the memory area to search in
[in]	type	Type of data (T_HTML or T_CSS)

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

◆ mobi_strip_mobitags()

MOBI_RET mobi_strip_mobitags ( MOBIPart * part )

Strip unneeded tags from html. Currently only <aid>

Parameters

[in,out] part MOBIPart structure

Returns: MOBI_RET status code (on success MOBI_SUCCESS)

Macros

Functions

Detailed Description

Function Documentation

◆ mobi_embed_to_link()

◆ mobi_find_attrname()

◆ mobi_find_attrvalue()

◆ mobi_flow_to_link()

◆ mobi_get_aid_by_offset()

◆ mobi_get_aid_by_posoff()

◆ mobi_get_aid_offset()

◆ mobi_get_attribute_value()

◆ mobi_get_filepos_array()

◆ mobi_get_id_by_offset()

◆ mobi_get_id_by_posoff()

◆ mobi_get_ncx_filepos_array()

◆ mobi_get_offset_by_posoff()

◆ mobi_get_rawlink_location()

◆ mobi_iterate_txtparts()

◆ mobi_markup_to_utf8()

◆ mobi_posfid_to_link()

◆ mobi_process_replica()

◆ mobi_reconstruct_flow()

◆ mobi_reconstruct_infl()

◆ mobi_reconstruct_infl_v1()

◆ mobi_reconstruct_links()

◆ mobi_reconstruct_links_kf7()

◆ mobi_reconstruct_links_kf8()

◆ mobi_reconstruct_orth()

◆ mobi_reconstruct_parts()

◆ mobi_reconstruct_resources()

◆ mobi_search_links_kf7()

◆ mobi_search_links_kf8()

◆ mobi_strip_mobitags()