[Lxr-dev] Re: [PATCH] Preliminary GIT support

[Malcolm B. <mal...@gm...>]

Jan-Benedict Glaw wrote:
> On Fri, 2006-03-31 23:17:00 +0100, Malcolm Box <mal...@gm...> wrote:
> 
>>$output = `command`;
>>if($? >> 8) {die "Problem with command";}
>>
>>if you want to get rid of the \n on the string, then do:
>>
>>chomp $output;
> 
> 
> Is there a way so supply the arguments one-by-one? I'd like to get
> things like whitespace in a path correct, thus I guess something like
> exec() vs. system() may be needed?

You can use system() with a list of parameters, which will then preserve 
whitespace in pathnames (evil though this is). e.g. 
system("git-command", "param1", "path with spaces in it")

This doesn't capture the output though, so you might have to fork and 
read the output.  There's a good example of how to do this in the 
perlsec man page - look for "backtick".  Doing things that way would be 
a good idea if the command is ever run in response to untrusted user 
input - i.e. if it can be triggered by the browser interface, since 
there would then be a potential injection attack.

The other thing you could try is quoting the pathnames in the backticks 
- from some simple experiments this would appear to work.

> 
> 
>>> * Given that a program outputs numerous lines of which the first
>>>   lines have a strict syntax (and the following are free-form text),
>>>   what's the easiest way to get one of the lines with known format
>>$output = `git-cat-file commit...`;
>>if ($? >> 8) { die "Something interesting happened";};
>>
>>($author_line) = ($output=~m/^(author .*)$/m;
>>($author_email) = ($author_line=~/<(.*)>/;
>>
>>or something like that should do the trick.  The basic flow is to read 
>>everything into the $output variable, then use the regexp operators of 
>>perl to chop out what you need.
> 
> 
> Is the unmatching number of opening vs. closing braces intended
> (right-hand)?

No, sorry!  Comes of posting code that I've not run.  It should be:

($author_line) = ($output=~m/^(author .*)$/m);

>>The first regexp uses the /m modifier so that it will deal with a 
>>multi-line string in $output.  The second doesn't need this since we've 
> 
> 
> Does this take care of throwing away all but the first author line?

The $author_line will contain only the first line starting with author. 
  So it does throw away all the rest.

>>just chopped out one line.  Both regexps use the fact that in list 
>>context a match (m//) operation returns a list of all matched bracketed 
>>expressions.

Cheers,

Malcolm