|
|
This file describes the format of the dump file. The data stored in the dumpfile are dynamic, depending on
the flags you have set in the DumpingInfo class.
Using the dump file can speed up the loading process. Especially if the CHM file contains a large sitemap table
of contents/index file (~800KB may take up to 10 sec on 1.8GHz CPUs). Switching from the sitemap-format
to the dump file can speed up the loading process over 90% (same toc read from dump in ~<0.5 sec on 1.8GHz CPUs).
If your CHM contains a large binary index/toc the dump-file usage aprox. halves the load time.
If you use dumping for files with a large amount of index/toc entries, it may be performant if you add the #URLSTR
and #STRINGS files to the dump (using the dumping flags).
Examples:
Test done on: Intel Xeon 3.06GHz HT, 1GB Ram, WinXP, SCSI-HDs
DumpingFlags set: DumpingFlags.DumpBinaryTOC | DumpingFlags.DumpTextTOC |
DumpingFlags.DumpTextIndex | DumpingFlags.DumpBinaryIndex |
DumpingFlags.DumpUrlStr | DumpingFlags.DumpStrings
Dump comrpession: DumpCompression.Medium
DirectX9 SDK CHM (binary index and binary TOC):
Read time without dump: --- HtmlHelp file read in 00:00:02.1874784
Write time of dump data*: --- Dump written in 00:00:01.0937360 (dump file size: ~780KB)
Read time with dump: --- HtmlHelp file read in 00:00:00.7499904 (dump file size: ~780KB)
Net read time of dump: --- Dump read in 00:00:00.5781176 (dump file size: ~780KB)
CHM with a ~900KB sitemap TOC (binary index and text-based TOC):
Read time without dump: --- HtmlHelp file read in 00:00:05.7811760 (slow RegEx parsing :( )
Write time of dump data*: --- Dump written in 00:00:00.5156184 (dump file size: ~300KB)
Read time with dump: --- HtmlHelp file read in 00:00:00.3437456 (dump file size: ~300KB)
Net read time of dump: --- Dump read in 00:00:00.2187472 (dump file size: ~300KB)
* the write time of the dump is not included in the "Read time without dump" timespan.
I think the two examples above shows how the usage of the data dumping can speed up the loading process.
Another pro of using dump files is, that you will save a few MBs of memory, because only the necessary fields
are stored and loaded from the dump file. Also the initial size of strings is known when reading the dump,
so the .NET Framework can instantiate the string instance with an initial size.
The dumpfile starts with the following header:
BYTE size of signature text (n)
BYTEs n Bytes which forms a signature string
DWORD compression level
compression level description
------------------------------------------------------------------
0 None, no compression stream is used
1 Minimum compression
2 Medium compression
3 Maximum compression
Depending on the compression level, the following block differs (see ChmDecoing\DumpingInfo.cs line 312
(for writing) and line 369 (for reading)).
If compression level = 0 (no compression), the following format can be read 1:1 from the
dump file. If compression level > 0, you have to create an InflaterInputStream() instance
from the ICSharpCode.SharpZibLib library and attach it to the current file stream.
Using this inflater stream, you can read the compressed data as described below (decompression
will be done by the inflater):
BYTE size of signature text (n) (same as in header above !)
BYTEs n Bytes which forms a signature string (same as above)
BYTE size of timestamp string (m)
BYTEs m Bytes which forms a date/time for the last write access of the CHM
The string has the following format: "dd.MM.yyyy HH:mm:ss.ffffff"
DWORD used encoding ansi-codepage
BYTE boolean flag (0=false, 1=true) specifying if the dump should contain the data
of the #STRINGS file.
if the previous flag returns TRUE
BYTE boolean flag (0=false, 1=true) specifying if the #STRINGS data is supported by the CHM
if the previous flag returns TRUE the dump of #STRINGS follows
DWORD number of dictionary pairs (key, value) (cnt)
for each dictionary entry (cnt times)
DWORD offset of the string entry in the #STRINGS file (=key)
BYTE size of following string (n)
BYTEs n Bytes which forms the value string
BYTE boolean flag (0=false, 1=true) specifying if the dump should contain the data
of the #URLSTR file.
if the previous flag returns TRUE
BYTE boolean flag (0=false, 1=true) specifying if the #URLSTR data is supported by the CHM
if the previous flag returns TRUE the dump of #URLSTR follows
DWORD number of dictionary pairs for urls (key, value) (cnt)
for each dictionary entry (cnt times)
DWORD offset of the string entry in the #STRINGS file (=key)
BYTE size of following string (n)
BYTEs n Bytes which forms the value string
DWORD number of dictionary pairs for frame names (key, value) (cnt)
for each dictionary entry (cnt times)
DWORD offset of the string entry in the #STRINGS file (=key)
BYTE size of following string (n)
BYTEs n Bytes which forms the value string
BYTE boolean flag (0=false, 1=true) specifying if the dump should contain the data
of the #URLTBL file.
if the previous flag returns TRUE
BYTE boolean flag (0=false, 1=true) specifying if the #URLTBL data is supported by the CHM
if the previous flag returns TRUE the dump of #URLTBL follows
DWORD number of urltable entries (cnt)
for each urltable entry (cnt times)
DWORD offset into urlstr file
DWORD offset of this entry
DWORD index into topics file
DWORD offset into urlstr file
BYTE boolean flag (0=false, 1=true) specifying if the dump should contain the data
of the #TOPICS file.
if the previous flag returns TRUE
BYTE boolean flag (0=false, 1=true) specifying if the #TOPICS data is supported by the CHM
if the previous flag returns TRUE the dump of #TOPICS follows
DWORD number of topic entries (cnt)
for each topic entry (cnt times)
DWORD offset of the entry
DWORD offset into tocidx file (binary toc)
DWORD offset into strings for the topic title
DWORD offset into urltable
DWORD visibility mode
DWORD unknown mode
BYTE boolean flag (0=false, 1=true) specifying if the dump should contain the data
of the $FIftiMain (Full-text search) file.
if the previous flag returns TRUE
BYTE boolean flag (0=false, 1=true) specifying if the $FIftiMain data is supported by the CHM
if the previous flag returns TRUE the dump of $FIftiMain follows
Header of full-text engine
DWORD number of index files
DWORD root offset
DWORD page count
DWORD depth of the tree
DWORD scale for document index
DWORD root for document index
DWORD scale for code count
DWORD root for code count
DWORD scale for location codes
DWORD root for location codes
DWORD size of the index/leaf nodes
DWORD length of longest word
DWORD total number of words
DWORD total number of unique words
End of header
DWORD number of bytes following (binary full-text index) (n)
BYTEs n Bytes which represent the binary full-text index byte array
BYTE boolean flag (0=false, 1=true) specifying if a table of contents is in the dump file
if previous flag returns TRUE
DWORD number of TOC items (n)
A: for each toc item (n times)
DWORD toc mode (0 = text based toc, 1 = binary toc)
DWORD offset into topics file
BYTE size of the following string
BYTEs string which forms the name of the toc item
if toc mode = text based and topics offset < 0
BYTE size of the following string
BYTEs string which forms the local of the toc item
DWORD image index of this toc item
BYTE size of the following string
BYTEs string which forms a merge link (e.g. xy.chm::/toc.hhc)
DWORD number of information type associations (o)
for each information type association (o times)
BYTE size of following string
BYTEs string which forms the information type string
DWORD number of child toc items (m)
for each child toc item (m times)
each child item is a toc item, so repeat reading from mark (A:)
BYTE boolean flag (0=false, 1=true) specifying if a index is in the dump file
if previous flag returns TRUE
DWORD number of index items (ALinks) (n)
for each index item (n times)
BYTE size of following string
BYTEs bytes which forms the keyword string
BYTE boolean flag, true if the index item is a see also keyword
DWORD index indent
DWORD number of information type associations (o)
for each information type association (o times)
BYTE size of following string
BYTEs string which forms the information type string
DWORD number of see-also strings (cnt)
for each see-also string (cnt times)
BYTE size of the string
BYTEs bytes forming a see-also keyword
DWORD number of topic entries (cnt)
for each topic entry (cnt times)
DWORD topic mode (0...text-based, 1...binary)
if topic mode = 0
BYTE size of following string
BYTEs bytes forming the title string
BYTE size of following string
BYTEs bytes forming the topic local
if topic mode = 1
DWORD offset into topics file
DWORD number of information types in dump (n)
for each information type (n times)
DWORD information type mode
BYTE size of following string
BYTEs string wich represents the name of the information type
BYTE size of following string
BYTEs string which represents the descriptin of the information type
DWORD number of categories in dump (n)
for each category (n times)
BYTE size of following string
BYTEs string wich represents the name of the category
BYTE size of following string
BYTEs string which represents the descriptin of the category
DWORD number of information types assigned to this category (m)
for each information type (m times)
BYTE size of following string
BYTEs string wich represents the name of the information type
|
By viewing downloads associated with this article you agree to the Terms of Service and the article's licence.
If a file you wish to view isn't highlighted, and is a text file (not binary), please
let us know and we'll add colourisation support for it.
This member has not yet provided a Biography. Assume it's interesting and varied, and probably something to do with programming.
Submit an article or tip