Page MenuHomePhabricator

Support for rev_deleted in sql dumps
Closed, ResolvedPublic

Description

Author: Bryan.TongMinh

Description:
Patch for rev_deleted

Attached patch for the backup system to support the rev_deleted system.


Version: unspecified
Severity: minor

attachment worker.py-rev_deleted.patch ignored as obsolete

Details

Reference
bz13930

Event Timeline

bzimport raised the priority of this task from to Medium.Nov 21 2014, 10:10 PM
bzimport set Reference to bz13930.
bzimport added a subscriber: Unknown Object (MLST).

Bryan.TongMinh wrote:

Patch that actually outputs correct shell commands

Attached:

This doesn't really look suitable; entire records are being suppressed instead of the individual hidden fields.

For instance, we shouldn't drop an entire revision's metadata just because the comment was suppressed (which should blank only one field) or the text (which would not even be included in the output!)

Further, giving the same filename to dumps which contain different data looks unwise.

Bryan.TongMinh wrote:

(In reply to comment #2)

This doesn't really look suitable; entire records are being suppressed instead
of the individual hidden fields.

For instance, we shouldn't drop an entire revision's metadata just because the
comment was suppressed (which should blank only one field) or the text (which
would not even be included in the output!)

Well in that case we probably have to make our own version of mysqldump. I don't see how to do it otherwise.

Further, giving the same filename to dumps which contain different data looks
unwise.

Are you referring to the fact that the public and private versions of tables have the same name? But they go into different directories, right?

There's not a pressing need to do a public dump of the revision table these days, as the XML stub dumps contain that data.

Something nice for the logging table might be needed.

They go in the same directory for private wikis. Even if they didn't, having the identical name for *opposite data* is very bad and will be extremely confusing to anyone having to work with or preserve those files.

They do not go in different directories for private wikis.

So we need an xml log dump then?