[Qgis-community-team] Redmine -> Github tests

[Richard Duivenvoorde]

Hi Lists (cross posting to have both community, psc and developers),

Still busy ways to migrate from Redmine to Github possibilities, we play
with some options. Wanted to sent some ideas to the lists of current status:

Chatting to Matthias (who did a first import script based on Redmine
issue.xml imports [0][1]).

Chatting to Juergen (who did an import script based on table exports of
the Redmine database [2]).

Who both hit the Github-api-throtteling limit from Github [3] and
stopped there:
- max 60 api-search queries per minute
- max 5000 requests per hour for other api calls

We had a little chat to github support, we were pointed to a different
api [4]

Receiving python script from Juergen, and his plan to get as much
information from Redmine as possible by:
- creating a (sqlite) redmine.db with dumps of nessecary tables from
(mysql) redmine.db
- creating 'lookup'-tables for different mappings, eg map Redmine-users
toGithub-users. Have the Category-strings mapped to their id's AND
mapped to Github-labels etc etc.
- write a python script which queries the db and (via github api)
creates issues with exact the same issue-numbers as we have in Redmine
- add labels to a Github test repo as 'mapping to Redmine' things like:
status, prio, etc etc

My conclusion: if I/we invest enough time into this, I can do further
scripting and just import the issues in blocks, just enough to not hit
the limit.

I hand crafted 2 issues:

https://github.com/rduivenvoorde/temp/issues/17

https://github.com/rduivenvoorde/temp/issues/18

based on http://hub.qgis.org/issues/10755

To show/see some possible options:

- one Redmine comment = one Github comment (as in #17)
- all Redmine comments in one Github comment (as in #18)

You can see all possible labels if you try to create a new issue in the
repo:
https://github.com/rduivenvoorde/temp/issues

I tried to minimize the number of Labels, but tried to mimic most of
Redmine's.
By adding a number in front of it I hope we can instruct issuers to add
for example at least number 1 to 5 labels. And they are ordered by that
number in the Github views, making it maybe easier to 'read' them in an
apparent order.

All imports will be done by a 'QGIS-redmine-import' user.
All attributions will be done based on (Redmine) Name and (if available)
Github userid (I thin we should leave email addresses away)
References to commits and pr's can be mapped.
Assignee's can be set on Github userid's
Attachment info can be copied (url's to original attachment) BUT(!) the
api does not provide a way to add those. This can only be done via
Webinterface...

The plan is to take a couple of complex issues and fine-tune a script
till everybody is satisfied with result.

To do this work I need:

1) full backup of community that this is the way to go
2) help / feedback on first tests or coding
3) some time/days? (after 2.16 packaging: 8th july?) to freeze Redmine
and dump tables again
4) some encouragement

Already done: taking the 4079 users of Redmine, and checked with Github
if they have a user with the same email OR userid... Only 321 matched...

So also needed (separate threads/emails?):
- everybody check this list and see if he/she can match him/herself and
sent me his/her right Github id together with the Redmine one
- input/comments about current Labels
- if we miss something if we do it like this


Not sure how to channel this, but thought to sent it like an email

Regards,

Richard Duivenvoorde


[0] https://github.com/m-kuhn/rm2gh
[1] https://github.com/m-kuhn/QGIS-Test/issues/
[2] https://github.com/jef-n/rm2gh/issues/
[3] https://developer.github.com/v3/#rate-limiting
[4] https://gist.github.com/jonmagic/5282384165e0f86ef105