Review: Regular Expression Library

From: Beman Dawes (beman_at_[hidden])
Date: 2000-08-16 10:55:53

Conclusion

* This is a really neat library and should be accepted for boost.

* I reviewed regex by reading the documentation, building the library for 
one compiler, compiling and testing some of the sample code, and writing 
some of my own code using regex. Everything worked, with only trivial 
issues appearing (See below the line).

* I did only minor inspection of the regex source code itself.  Hopefully 
someone else will inspect the code in detail.

* Regex predates boost, and although boostified in key respects, still uses 
its own configuration and library build mechanisms.  Those should be 
migrated to boost's way of doing things over time, and as boost's library 
build mechanisms are developed.

-----------------------------------------------------------------------

Minor Issues

Design:

*  Having a class named "reg_match" and a function named "regex_match" 
caused some confusion.  Consider renaming reg_match to regex_match_results, 
match_results, or similar.

Documentation:

* The documentation would be more inviting to new readers if it began with 
a simple tutorial paragraph or two, before diving into the reference 
material.  A "hello world" level example would suck people in and get them 
interested.

* The code snippets show the power of regex.  They should be commented a 
bit more and made accessible directly from index.htm.

Nits:

*  vc6.bat:  Had to add -I to nmake command so del command failures 
wouldn't cause termination.  This may be a Win2K issue.

*  readme.txt apparently hasn't been boostified.

*  index.htm: The link to Demo Apps actually links to Thread Safety!

*  index.htm: Need a link to faq.htm

*  index.htm: Should there be a link to the traits class reference page?

*  Most of the mentions of files in the docs should be hyperlinked to those 
files.

Code:

*  Add title lines and // see www.boost.org... lines to source files per 
guidelines.  (Don't blame John for not following this part of the 
guidelines; they have only just appeared.)

*  Use of _MSC_VER to identify the Microsoft compiler.

Boost (rather than regex) issues:

*  We really, really, need a build/make process for separately compiled 
boost libraries.  Regex will make this need even more acute.

*  "identifier truncated to 255 characters" and similar useless warnings on 
lib build compiles. This is general boost irritation; we ought to figure a 
way to avoid it.

Thanks to John Maddock for regex!!

--Beman