Last minute cleanup of your working tree for apache projects

Minor patch updates don't require an IDE, nor

When commiting code to an apache project you want to follow guidelines like:

• All code is under 80 columns
• No trailing white spaces
• ...

Although IDE's can help you to prevent these errors, the simplest catch all solution is to maintain a custom shell script that will analyze your code for you in the changed files.

You can use git status to grab all the recent changed files.. And then use your terminal to run arbitrary analysis of your changed source files.

You can use git to find all the files you changed:

 git status --porcelain | grep java | cut -d' ' -f 3

This will list all your java files that changed, but  not commited yet. 

hadoop-common-project/hadoop-common/src/test/java/org/apache/hadoop/fs/TestPath.java

hadoop-common-project/hadoop-common/src/test/java/org/apache/hadoop/fs/TestStat.java

Then you can pipe this into egrep:

xargs egrep ".* +$"

Which will then list all offending files.

The final recipe looks something like this:

for f in `git status --porcelain | grep java | cut -d' ' -f 3 `
do
  
   echo "Processing $f.."
   cat $f | awk ' { if ( length > 80 ) { x = length } }END{ print x } '

   cat $f | xargs egrep ".* +$"
   echo "..Done processing $f" 
done

 Now, here's an important trick for editing a patch after you find out its broken.

- Put your code back to a state where it was BEFORE you wrote the patch. For example, you might do :

git checkout -- .  (if you hadnt commited the code in your patch)
git reset HEAD~1 (if you're whole patch is in the latest commit)

- apply the original patch

git checkout -- .
patch -p0 < ./HADOOP-3679.4.patch
- use git add -p , and hit "y" for all the diffs to keep, "e" or "n" to edit or nix them

git add -p
git diff ---cached --no-prefix > HADOOP-3679.patch