www.nico.schottelius.org/blog/find-emails-in-git-log-for-notification.mdwn

[[!meta title="Find e-mail adresses of people in git log output"]]

## Motivation

Some days ago I've replaced **cronwrapper**, a script to monitor output of
cron scripts with the replacement **cwrap** in local.ch's puppet configuration.

If the script prints on stdout, **cwrap** does not raise an error by default,
which **cronwrapper** did.

To notify every user of the change, I want to send an email to every 
ex-**cronwrapper** user.

## Solution

The configuration is stored in a subversion repo, which I locally sync using
**git svn**. Thus I can use **git log -p** to see all changes.

A typical line of interest looks like this:

    -        command => '/usr/local/bin/cronwrapper.sh EMAIL@EXAMPLE.COM "[mob][low][dev03-sth][front] description" /usr/bin/php /some/script',

Thanks to git, grep, sed, awk, there is a pretty simple solution 
(not the most beautiful) to this problem. First of all, get all patches:

    git log -p

Then find all removal entries of cronwrapper:

    grep ^- | grep cronwrapper

But only those containing an e-mail address:

    grep '@'

And filter out the e-mail address:

    sed 's/.* \(.*@.*\)/\1/' | awk '{ print $1 }'

Replace all quotes and backslash quotes:

    sed -e 's/\\"//g' -e 's/"//g' -e "s/'//g"

The problem now is that some e-mail adresses are indeed multiple e-mail adresses
(abc@example.com;def@example.com) and some e-mail adresses are lower, some upper case.

Breaking up the concatenated addresses can be done use awk easily:

    awk '{ gsub(";", "\n"); print $0 }'

Transforming all addresses to lower case can be done using the fine utility **tr**:

    tr '[A-Z]' '[a-z]'

Filter out all duplicates:

    sort | uniq

The result is a list of e-mail addresses. Making them usable for copy & paste
into webmail of exchange needs another filter to convert **\n** to **;**, but
add one **\n** at the end:

    awk 'ORS=";" { print $0 } END { ORS="\n"; print "" }' 

So in the end, the complete chanin looks like this:

    git log -p | grep ^- | grep cronwrapper | \
    grep '@' | sed 's/.* \(.*@.*\)/\1/' | awk '{ print $1 }' | \
    sed -e 's/\\"//g' -e 's/"//g' -e "s/'//g" | \
    tr '[A-Z]' '[a-z]' | \
    awk '{ gsub(";", "\n"); print $0 }' | \
    sort | uniq | \
    awk 'ORS=";" { print $0 } END { ORS="\n"; print "" }'

For me, this is a nice demonstration of the power of shell, unix tools and filtering via pipes.

[[!tag config sysadmin localch unix]]