regex status report #13

[Gabor Kovesdan]

Hi,

* I added support for some flags that has been forgotten.
* I added comments to the code.
* I added public interfaces to the fast matching code.
* I fixed several minor bugs and style and portability nits.
* I backported the fast matching code to BSD grep. It only gives a very 
small performance enhancement at the moment but this code is much better 
tested and fixes several bugs. I will commit this new version to ports soon.
* I added the foundation of the heuristic matching. At the moment, it 
has minor bugs and only supports BRE but the basic idea seems to work.

Looking at the original goals I can summarize the following:
* TRE builds as a part of libc and seems to havea good standard 
compliance, however when used out of the box, the performance is not 
always satisfying.
* The fixed string code is finished. There may be more optimization 
tricks and nits but basically it is finished.
* TRE has interfaces with byte-counted buffers to avoid reading the 
whole pattern/input, so it was not necessary to work on this.
* The foundations of a heuristic matching is ready but there is still 
work to do on this.
* No progress on GNU-specific regexes because higher priority tasks are 
still incomplete.
* Documentation (man page) is available from the old regex code but it 
does not describe TRE-specific interfaces, so it needs some more work.
* A regression test has been written and it is believed to be quite 
complete for basic cases.

I think it was a nice experience and I'm still working on this project. 
I hope we can see the deliverables soon.

Gabor