...?

Ever came across a dumb program which keeps running the same (slow) command over and over, expecting different output, but you know that the output couldn't possibly ever change?

Place a symlink to cache-wrap early in your $PATH, give it the name of the command you want to cache, and then just call the command as usual. cache-wrap will run the next command in $PATH that has the same name as the symlink, and stick the stdout into memcache. The next time the same command is ran from the same directory with the same arguments, the program will not run and instead the result will come from memcache.

examples

$ # (the script is in ~/cache-path/cache-wrap
$ ALT_PATH=$HOME/cache-path
$ ln -s ~/cache-path/cache-wrap $ALT_PATH/date
$ date
[...]
$ date
[same output as above, from memcache]

Note: if the command writes anything to STDERR then its STDOUT will NOT be cached.

options

You can set the environment variable CACHEWRAP_OPTS to a comma-separated list of options:

• help print this text and exit
• debug output some debugging information to stderr
• debug=file output some debugging information to file
• server=... specify Memcache server address (default 127.0.0.1:11211)
• filter=... Perl regular expression to decide whether to use the cache. The list of command line arguments is joined with spaces, and the result is matched against this regex. If it doesn't match, cache is disabled (as if the nocache option was used)
• nocache disable cache, always run actual command
• clear clear cache (delete from Memcache) then run the actual command
• nuke alias for clear

using options

$ CACHEWRAP_OPTS=server=2.3.4.5:11211,debug slow-cmd arg1 arg2
$ $OPTS_ENVVAR_NAME=filter='^date +%s$' slow-cmd arg1 arg2
# when slow-cmd calls 'date +%s', it gets a cached result, while
# any other date(1) invocation bypasses the cache