The Silver Searcher

My attempt to make something better than ack, which itself is better than grep. Right now it's nowhere close.

Why use ag?

• The binary name is 33% shorter than ack!
• It searches through code about 3x-5x faster than ack.
• It's written in C instead of perl.
• It ignores files matched by patterns in your .gitignore and .hgignore.
• If there are files in your source repo that you don't want to search, just add their patterns to a .agignore file. *cough* extern *cough*

Current development status

On the continuum of...

1. Compiles
2. Runs
3. Behaves correctly
4. Behaves correctly and runs fast

...it's somewhere between 2 and 3 right now. Although it's much faster than ack in my benchmarks.

ack -i SOLR ~/cloudkick/reach  2.89s user 0.77s system 97% cpu 3.765 total

ag -i SOLR ~/cloudkick/reach  0.37s user 0.15s system 88% cpu 0.590 total

You can use this with my fork of the popular ackmate plugin, which lets you use both ack and ag for searching in textmate. You can also just move or delete "~/Library/Application Support/TextMate/PlugIns/AckMate.tmplugin/Contents/Resources/ackmate_ack" and run ln -s /usr/local/bin/ag "~/Library/Application Support/TextMate/PlugIns/AckMate.tmplugin/Contents/Resources/ackmate_ack"

TODO

• parse more ack options correctly
• actually get textmate dir patterns working
• support inverted matching
• support piping in stuff
• docs/manpages

Other stuff you might like

• AckMate
• Exuberant Ctags
• Git-grep

initial brainstorming notes for myself

• ack annoyances:

  • only uses one core
  • can't prioritize certain dirs/files
  • can't say "just ignore stuff that's in .gitignore/.svnignore/.hgignore/whatever"
  • can't easily introspect what files are being ignored/searched
  • poor/no support for searching/ignoring files with no dots in their name or dots at the beginning or multiple dots, etc
  • ack is a giant hack. it's full of crazy stuff that perl programmers do. tons of special-cases and other dumb stuff
  • ack hates anything that isn't utf-8
  • by default, it hardly searches anything. you have to edit your ~/.ackrc to get useful behavior * grep is on the opposite side of the spectrum
  • no easy way to show skipped files in a directory hierarchy
  • if there are no matches, ack doesn't automatically try some less accurate searches * concrete example: if no matches, enable ignore case and redo the search
  • basic options: * follow/don't follow symlinks * recurse dirs

• ack awesomeness (aka stuff I want to copy):

  • can easily skip files (unlike grep)
  • faster than grep
  • integrates with textmate (ackmate)

• code search: common actions

  • find vars or functions with same/similar names in the project

• pie-in-the-sky stuff

  • integrate with filetype-specific libraries (libxml2 etc) for extra functionality
    • concrete example: searching some xml files
      • instead of printing the context lines (prev/next 5 lines), print everything in the local dom scope
    • track code dependencies (this is almost certainly a different product/program)
  • switch to RE2 http://code.google.com/p/re2/ http://john.freml.in/re2-benchmark
  • introspection
    • why did or didn't you search this file? (because of .agignore or hardcoded paths or it's a symlink or what?)
    • how many files searched in how much time? average MB/sec