Refine audit:code ordering

[joshpencheon]

The intention is that show_in_priority=true can be passed to rake audit:code, which should output files ordered such that those which have the oldest unreviewed code appear first.

However, I'm not sure that it is working as intended. Current implementation:

https://github.com/PublicHealthEngland/ndr_dev_support/blob/master/lib/tasks/audit_code.rake#L87-L92

We should be ordering files by:

• if the file has not been reviewed before, the revision the file was added to the repository, or
• if the file has been reviewed before, the revision that the file was last flagged as safe (safe_revision), and not the last_changed_rev (which might be misleadingly-recent for files that are touched regularly).

This may still not be perfect.

In [git-]svn, revisions are numeric and therefore orderable. In pure git, commits are identified by a digest. Some derivable orderings:

# returns 'n' , when <<SHA>> is for the n-th commit on the current branch
$ git rev-list --count <<SHA>>

# returns the commit date of <<SHA>>
$ git show -s --format=%ci HEAD~10

Other potentially-useful commands:

# In git-svn, the revision that added <<FILE>>:
$ git svn log --oneline --reverse --limit 1 <<FILE>> | cut -d\| -f1

# In svn, the revision that added <<FILE>>:
$ svn log -r 1:HEAD --limit 1 -q <<FILE>> | grep -v '^-' | cut -d' ' -f1

[timgentry]

@joshpencheon this would be much easier to implement if we switch to using Rugged, but that would mean we would have to drop support for native subversion code review, as it would have to be done using git-svn. Thoughts?

[joshpencheon]

Cool - it's not something I'd previously considered.

I do have a (stale) branch with efforts to make this work on; I think it was just painfully slow for larger repositories:

https://github.com/PublicHealthEngland/ndr_dev_support/tree/audit_priority

[joshpencheon]

@bshand just making you aware of this (rather old) issue, with respect to our recent conversation about code review.