ecometrica / grandfatherson Goto Github PK

Hi,

First, thanks for developing this grandfatherson! I use it as a requirement of my project: bakthat, I there is an issue with the firstweekday keyword argument, here is an example: http://nbviewer.ipython.org/5553570

With first firstweekday=grandfatherson.MONDAY:

('Monday', datetime.datetime(2013, 3, 4, 0, 8, 4, 4437)),
 ('Saturday', datetime.datetime(2013, 3, 9, 0, 8, 4, 4437)),
 ('Saturday', datetime.datetime(2013, 3, 16, 0, 8, 4, 4437)),
 ('Saturday', datetime.datetime(2013, 3, 23, 0, 8, 4, 4437)),
 ('Saturday', datetime.datetime(2013, 3, 30, 0, 8, 4, 4437)),
 ('Saturday', datetime.datetime(2013, 4, 6, 0, 8, 4, 4437)),
 ('Saturday', datetime.datetime(2013, 4, 13, 0, 8, 4, 4437)),
 ('Saturday', datetime.datetime(2013, 4, 20, 0, 8, 4, 4437)),
 ('Saturday', datetime.datetime(2013, 4, 27, 0, 8, 4, 4437)),
 ('Saturday', datetime.datetime(2013, 5, 4, 0, 8, 4, 4437)),

Only the first weekly backup is kept on Monday, Saturday for the others.

Thanks!

• It could be clarified whether calculations are inclusive or exclusive of the current date of 'now'. The answer appears to be inclusive. It coudl be worth mentioning that this means that this tool should run after daily snapshotting happens.
• It could also be clarified that "days:3" does NOT mean the "the 3 most recent daily backups", it means "backups from the 3 most recent days". If you have 3 three daily backups but none in the last 3 days, this tool would select all your backups for deletion and leave you with zero. There's no problem as long as the backups are running fine. But if you are backups get stalled and your pruning script does not, you'll slowing eat through your backups.

When using Python 3 instead of Python 2.7 the following error appears when invoking the grandfatherson.dates_to_delete function:

File "/usr/local/lib/python3.4/dist-packages/grandfatherson/__init__.py", line 220, in dates_to_delete
    firstweekday=firstweekday, now=now)
  File "/usr/local/lib/python3.4/dist-packages/grandfatherson/__init__.py", line 204, in dates_to_keep
    firstweekday=firstweekday, now=now)
  File "/usr/local/lib/python3.4/dist-packages/grandfatherson/__init__.py", line 172, in to_keep
    filters.Minutes.filter(datetimes, number=minutes, now=now) |
  File "/usr/local/lib/python3.4/dist-packages/grandfatherson/filters.py", line 55, in filter
    if number < 0 or not isinstance(number, (int, long)):
NameError: name 'long' is not defined

The current recommendation when migrating from Python 2.7 to Python 3 is to use an int instead of a long.

In #6 I recommended documenting a data loss case that might be surprising: If your backups are interrupted, a pruning tool based on grandfatherson could delete more backups than you might like because days:3 deletes backups older than 3 days ago and does NOT attempt to keep 3 days worth of backups.

That data loss case can be avoided by changing the default value of 'now' to match the most recent timestamp of the backups.

If that change were made, an interruption in daily backups being made would still result in days:3 retaining the most recent 3 days of backups that were successful.

I understand that data retention rules also call for making sure that backups are eventually deleted, but it seems better for defaults to err on the side of "data retention" then data loss!

For the case where daily backups are being made and the pruning tool runs after the daily backups, there would no difference in the result.

When your backups fail to run for some reason might be exactly the time that you want to preserve older backups in the series and not keep pruning the list smaller and smaller while backups are offline.