There may be other flags not related to list mode that don't make sense to have, too. -A, -B and -C are the ones I've found so far.

ack 1.x's behavior was to only search known files. ack 2.0 will make this not be the case. For people who like the 1.x way, make an --only-known option, especially for putting in your ~/.ackrc

I'm not set on the name, but we need the functionality.

Normally, ack gets its list of files to search from ARGV. This feature would allows uses to specify a file containing one filename per line that specifies the list of files.

If you write the following in Ack.pm:

{
my $first_time = 1;

sub call_func {
  if($first_time) {
    $first_time = 0;
  } else {
    print "again\n";
  }
}

}

and call call_func, you would probably expect no output from ack, but you'd be wrong. Squash puts .pm files after the call to main(), so variables initialized inline never get initialized. We either need to warn the developer if they make this mistake (and recommend they initialize their variable in a BEGIN block), or fix squash.

The ackrc needs to get installed into /etc/ackrc when ack installs.

Do we need to handle the case when there's already one there and not overwrite it?

Inverting a filter doesn't work.

Sometimes, as a user, you don't have write access to the /etc/ directory (ex. in a shared hosting environment). We should define semantics for an finding an alternate global ackrc should this be the case.

See GH #6.

We have no test for this!

It might be nice to allow a project file to specify that the user and system RC files should not be consulted.

All we have to do is implement context and passthru.

Tainting is important.

$ ack -g smtp
smtptunnel
$ ack -g SMTP
$ ack -g SMTP -i

This last case with -g -i should find smtptunnel as a filename.

If you're operating under a descendant of your home directory, and you have no project ackrc, ack2 will use $HOME/.ackrc for your project file as well as your user file. This results in $HOME/.ackrc being loaded twice.

I don't think we want to use .gitignore automatically, but maybe in ack 2.1 we can do something like add a --ignorefile=.gitignore.

Why can't I just say ack --perl 'sub foo' --pager? It should default to PAGER, in my opinion.

User @fayland wrote to ask in beyondgrep/ack1#222

is it possible to add --mtime supports for modified time (-M in Perl). usually I just want to search the files which modified today or last week.

These three results should be separated by a blank line between each

tags:19706:thisisasub   somenu2.php /^                            $thisisasub=$good_to_go=1;$/;"    v
Dev/xrefmap.txt:6182:$good_to_go
somenu2.php:88:                $good_to_go=0;

I use ack on a lot of different machines, so I think that having the option to have a default ackrc "compiled into" might be nice to ease distribution. Maybe even have an option for this, like:

ack --pack-ackrc=my-ackrc

Thoughts?

Usage: ack [OPTION]... PATTERN [FILE]

Even in usage, we say "file" even though it could be a directory. Make it obvious and clear that ack need not take an actual filename.

When I type ack --perl -g -i access, I expect output similar to find . -iname '*access*.pl' -or -iname '*access*.pm' -or -iname '*access*.t'. Should we add this to ack2? @petdance, thoughts?

When I try to search my source tree, I get the following error:

$ ack2 -l FOO
ack: .git: Permission denied
Can't call method "name" on an undefined value at d:/git/ack2/ack line 1247.

I was able to trace this back to App::Ack::Resource::Basic line 37, which attempts to open the file it's been passed for reading. If this file is a directory, then this attempt fails (yielding the "Permission denied" error) and the method returns no object (which explains the "Can't call method on undefined value" error immediately afterward). If I make a local modification to the module so line 37 reads

if ((! -d $self->{filename}) && !open( $self->{fh}, '<', $self->{filename} ) ) {

then I get the following output:

$ ack2 -l FOO
Doxyfile
lib/GA/MiddleTier/Ops/Sync/WizardData/Mssql.pm

More interested that it dumps everything, rather than caring about the formatting.

Currently, when I run the following:

ack --perl --ruby $awesome_stuff $the_internet

ack interprets this as (search all Perl and Ruby files in $the_internet for $awesome_stuff). So it means Perl OR Ruby. I was thinking about implementing #35 (and my extended idea in #34) as a filter; however, for that to work properly, I would need a way to say --mtime AND --perl, or --dirty AND --ruby.

The --ignore-dir option currently only supports direct matches (ex. --ignore-dir=is,extlib; --ignore-dir=extlib is just shorthand for that). In the future, it'd be nice to allow more complex match types.

Specifying --dump currently dumps out the options loaded from config files. It'd be nice if we had other --dump types (which makes me want to rename this option to --debug), such as:

• Filters - show a representation of which filters are selected by the other options on the command line.
• Filter Step - allow you to step through the execution of ack (like a debugger), and see the effect of each filter on each resource.

--dump dumps all the config options. Add a test for it.

Requires grouping and --output.

I'm getting this test failure:

t/config-finder.t .......... 1/21 
#   Failed test at t/config-finder.t line 69.
#     Structures begin differing at:
#          $got->[1] = '/home/shlomif/.ackrc'
#     $expected->[1] = Does not exist
# Looks like you failed 1 test of 21.

The problem is in App::Ack::ConfigFinder that also searches all upper directories for .ackrc files and list them and I have an .ackrc in my home directory.

Rob and Andy are making a lot of updates. Make a maketarget for them to install ack to ~/bin/ack2 and ackrc to ~/.ack2rc

• Define the rules for placement of /etc/ackrc on all platforms
• CPAN installer has to install /etc/ackrc
• ack --install-ackrc

See http://stackoverflow.com/questions/9625316/what-color-options-exist-for-ack-grep-for-colorization-of-output-logs-etc/9626444#9626444

In ack 1.x, I can write the following:

ack --type=perl -f # means the same as ack --perl -f

This doesn't currently exist in ack2, so we need to add it.

Allowing --noenv in a .ackrc file will cause crazy results. We have to check for --noenv before we go looking in the environment or the filesystem, and if we see it, that's an error.

There are big differences. Some of these differences may mess up ack 2.0 users, like when they find that their massive text files get searched that never were before.

Some options, like --dump, are currently not being checked for in our option coverage. We need to make sure that they are included.

Ben ([email protected]) emailed me about adding support for R in ack. I said I wasn't going to add anything to ack 1.x, but this sort of type handling is something we need to be aware of going forward. We already have C as --cc, but what will we do with R? D?

hello --

thanks for the great product.

1) i think R is being used enough to deserve a type in ack.  i think
.R is a common extension for it.

2) in a related note, i had

--type-set=R=.R

in my .ackrc but it didn't work.  i looked at the source and
discovered that this is because ack down-cases the file extensions
before comparing with the known extensions.  hence any extension
that is not all lowercase will never match.  changing it to

--type-set=R=.r

fixed it.  this is a little confusing for people who are used to
upper case extensions (as with R.)  as such i propose that ack
automatically internally down-case any extension specified by a
--type-set or --type-add, to prevent confusing users.

just my 2 cents.

best,
ben

run_ack and friends from Util.pm are how we run ack in a test environment, so these functions have access to all the options we actually test. It'd be smart to create a summary of which options are used, so we can verify that we're testing the program's functionality properly.

Many terms thrown around, need to understand them all

• "filetype"
• "resource"
• "rule"
• "selection"

In #35, it is proposed to add an --mtime option to ack. I think this is a great idea; in fact, I feel like the filter system built into ack2 is perfect for the job. That way, I could create the mtime filter in my ackrc, like so:

--type-add=mtime,command,find . -mtime $n_days

However, where does $n_days come from? I shouldn't have to hardcode this in the config file; otherwise I'd need --mtime1, --mtime5, --mtime10, etc. It'd be nice if you could just say --mtime=10, so I think the filter system should support this.

Currently, if you have an ackrc file in your home directory, the ack2 test suite will try to use it. This is bad.

We've started the ackrc, but it needs more filetypes brought over from ack.

Ok, you know the title of this issue? It's probably not that simple. Either way, I was building the Awesome window manager (http://awesome.naquadah.org/), and ack2 --cmake -f only listed CMakeLists.txt. I'll investigate later.

I think it might be nice to allow --passthru to throw a bell character into the match. Let's say I'm tailing a log file, and I want to highlight the string "Job's done":

tail -f /var/log/warcraft/peasant.log | ack --passthru --color "Job's Done"

If I expect that job to take a while, I may only check it every few minutes. However, if ack had the ability to throw a bell character into the match coloration, I would hear a bell every time that pattern is matched, and I would know to check that log file again.

See beyondgrep/ack1#220.