Re: Patch for transparent decompression

[Tony Abou-Assaleh]

Reuben Thomas wrote:
> > * In addition to documentation, I'd like to see some test cases. And 
> > the more the better :O)
>
> I'm not good at thinking of these; I'd be happy for any hints.

* Sample compressed file with a few common options passed to grep
* Pass through uncompressed file to test error handling
* Compile with the option disabled and test for error message
* Compile with the option enabled and test that it is in fact enabled
* A test with UTF-8 non-ASCII file name and file content would be good 
as well.

> > * My understanding is that zlib support .gz files. Any plans to 
> > integrate libbzip2 as well?
>
> This is a nasty one. I see three options:
>
> 1. [Easiest for grep] Add code to grep to determine what sort of 
> compression we're looking at. Currently we don't have to do this.
>
> 2. [A bit cleaner, a bit slower] Write a little library that decides 
> whether to delegate to libbzip2 or libz for a given stream.
>
> 3. [Nicest solution, slowest and most difficult to get accepted] Patch 
> libz so that it can delegate to libbzip2 for bzipped files (or vice 
> versa), and just use the one interface.

I was thinking more of providing 2 options, so the user can tell grep 
what kind of compression we use. But now that you mentioned it, we could 
add a parameter to decompress: gz, bz2, auto. Depending on how easy/hard 
auto is, we could start without auto and introduce auto later. We could 
also default to gz for now, and later default to auto without causing 
too much trouble.

Cheers,

TAA

--
Tony Abou-Assaleh
Email:    address@hidden
Web site: http://tony.abou-assaleh.net