This a stupid scheme for letting me jump around in source code since I can never seem to get TAGS files to work the way I want. It provides two minor modes, jumper-mode and jumper-line-mode. The former is useful when looking at source code and can be used to jump to the definition of the symbol at any point via the M-. key combo and back to where you came from with M-,. (The locations are arranged in a stack so if you M-. several times, you can unwind with repeated applications of M-,.)

The other mode, jumper-line-mode, is useful in shell buffers. It also binds M-., but this time to a function that tries to find the name of a file and, optionally, a line number in the current line and then jumps there. It currently understands error messages and stack traces from Perl, Python, and Ruby as well as the output of grep -nr. As in jumper-mode, M-, pops you back to where you came from and the two modes use the same stack so you can M-. to a file from jumper-line-mode and then explore further with the jumper-mode M-. and eventually M-, your way back to where you started.

To find a definition, jumper-mode looks in a file named JUMPER that lives somewhere above the source file you are in. Each line of this file should contain a name, file name, line-number triple delimited by tabs. The scripts rubydefs and scaladefs included with this library provide a quick and dirty way to do this for Ruby and Scala code. Here's how I generate my JUMPER file:

rm -f JUMPER
touch JUMPER
find . -name '*.rb' | xargs ~/hacks/jumper/rubydefs >> JUMPER
find . -name '*.scala' | xargs ~/hacks/jumper/scaladefs >> JUMPER

To install the elisp, place the file jumper.el in a directory that's in your load-path (or use add-to-list to add the directory it's in to load-path) and add these lines to your .emacs to enable this mode, for example, in Ruby and Scala code and in shell-mode.

(require 'jumper)
(add-hook 'ruby-mode-hook 'jumper-mode)
(add-hook 'scala-mode-hook 'jumper-mode)
(add-hook 'shell-mode-hook 'jumper-line-mode)

There's, at the moment, nothing particularly clever about this: if there are multiple definitions for the same name your get taken to some random one. I plan to fix that if it turns out to be a real problem in practice.