chm

Overview

The FPC CHM package is a set of units distributed with FPC 2.2.2+ that allow to read/write .chm compressed HTML help files on multiple platforms.

There are two projects that use the CHM package:

1. the textmode IDE (FPC 2.2.4+)
2. the Lazarus chmhelp package (which features lhelp, a separate help viewer, connected over TCPIP). The original author is Andrew Haines, but some parts are by others (specially Lars/z505).

The write aspect is mostly used in combination with fpdoc, a doxygen like documentation tool, and chmcmd the commandline compiler.

It is expected that in the next FPC release, chm will replace the doc-html archive with its thousands of separate htmls.

The package is pure Pascal, and thus portable in principle:

• The package is prepared for different endianness, but has not been extensively tested on Big endian systems
• No testing has been done on systems that generate an exception on unaligned memory access (like arm-wince). Specially the reader parts should be made aligned() clean.

The package also packages two examples:

1. chmls
2. chmcmd (see below)

Package status

The basic system works since 2.2.4. However keep in mind that the FPC help systems, while massive (30000+ webpages in the various chms) are generated, compiled and read mostly by FPC tools (with an occasional test in Windows/gnochm/kchmviewer to see if it opens properly).

If you want to use FPC's chm package/compiler for your own projects, take the follow advice under consideration:

• Work on data that is backed up. I haven't lost data yet, but better be safe than sorry.
• Don't expect it to work right away.
• Keep in mind what is implemented and what is not (see below)
• report bugs in the FPC tracker with a reproducible description (if you are in an hurry with patch :_)

Units

1. chmbase - some structures, constants and helper funcs (compare and compression streaming helpers)
2. chmfiftimain - the unit that is responsible for reading and writing the search index of chms (TChmSearchReader, TChmSearchWriter)
3. chmfilewriter - TChmProject, a class that describes a CHM project in the form of a XML description. Can be used to generate CHMs easily.
4. chmreader - the base CHM Reader
5. chmsitemap - Support for sitemaps (toc and index, the unit helps transforming these xml files to collections)
6. chmspecialfiles - Streaming helpers for special index files in the chm
7. chmtypes - CHM helper class types.
8. chmwriter - the base CHM Writer.
9. fasthtmlparser - Base skeleton of a html parser.
10. htmlindexer - Contains classes that TChmWriter uses to parse and index the contents of html files for searching
11. htmlutil - helperfunctions. Mostly getting info out of html/xml tags.
12. lzxcompressthread - wrapper around paslzxcomp that will do compression in multiple threads
13. paslznonslide - (de)compression routines
14. paslzxcomp - (de)compression routines
15. paslzx - (de)compression routines

Files

This describes mostly the FPC 2.6+ situation.

Besides the library units, there are also two end user programs:

• chmls - listing, extracting and unblocking (needed on Windows XP SP2+ and presumably newer Windows versions) of a chm
• chmcmd - creates a CHM using a XML file made with the TCHMProject class. Roughly equivalent to a simple help compiler. Experimental support for .hhp is also available.

chmcmd is, like the library itself, more a backend tool than a user friendly tool. It is unforgiving if your files are inconsistent (will throw exceptions on missing files), and does not support goodies like autoindexing. It is probably only a substitute for backend jobs where you ensure the input is consistent.

Missing features

• merged chm support/ #idxhdr
  • Combining of indexes of multiple chm files. (.CHW/.CHI?)
  • slave file support ? (making slave CHMs, and getting them out, will make for easy master/toc CHMs later)
• Writing of binary indexes is multiple level, loading not. (???)
• Expand TCHMProject to support
  • aliases. 2.4.2+ (via .hhp in 2.4.4+)
  • control over textual/binary toc generation, textual/binary index generation etc from the XML/project (2.4.2+)
  • also support .ini (MS helpcompiler compat?) (2.4.4+)
  • lcid/language (?)
  • Window list (in progress, probably 2.4.4+)
• Alinks
• auto index/toc(?) generation
• (popup) comments
• samples (fpdoc?) Make samples extractable, MSDN style.
• exclusion lists (stoplists) for wordindexing.
  • why is a MS generated index larger while we also index words like "the" and "of" ?

      Probably the parser doesn't see all text.

Implemented features

• Full text search
• Binary Table of contents
• Binary Index (experimental)
• Experimental threaded LZX Compression
• context ids.
• threaded compilation (requires recompile, see below)
• windows support
• toc and indexfile are not always "default.hh[c/k]"
• chmcmd and chmls totally revised, and chmcmd now also compiles .hhps.

Disable binary toc/index generation

Generating binary forms can be disabled in fpdoc respectively with --no-bintoc and --no-indextoc.

Note that these options were not added to save space, but to be able to temporarily not generate them if some viewer has a problem with them, and refuses to fall back to default.

One can notice that kchmviewer loads chms with binary index and toc significantly faster.

How to enable threaded LZX Compression

If you want to test multithreaded chm compression add -dLZX_USETHREADS when compiling the package (or the whole snapshot).

The program using the package (fpdoc must have cthreads added in its uses section on *nix.

On my Core2 1.73GHz laptop this reduced time from 2min 57s to 2m 7s, in other words a reduction by slightly less that a third.

It is unknown how to configure the number of parallel threads.

See also

• chm backend for fpdoc
• Textmode IDE development#Chm Support
• Lazarus chmhelp component

Go to back Packages List [[Category:Help and Docs]