Skip to content

Make date string regex more lenient

Joel Goguen requested to merge jgoguen/tracker:wip/jgoguen/date-time-format-parsing into master

The current regex, although technically exact according to the ISO8601 standard, gives messages such as this for any file with metadata added without using a zero-padded hour offset:

Could not insert metadata for item "file:///home/jgoguen/Music/Yoshikazu%20Mera,%20Japan%20Philharmonic%20Orchestra%20&%20Shigeo%20Genda/The%20100%20Most%20Essential%20Pieces%20of%20Classical%20Music/Xerxes,%20HWV%2040:%20%22Ombra%20mai%20fu%22%20(Largo).m4a": Not a ISO 8601 date string. Allowed form is [-]CCYY-MM-DDThh:mm:ss[Z|(+|-)hh:mm]

According to tracker extract there's only one date in the file metadata for this music file:

nie:contentCreated "1966-06-11T06:27:46-7:00"

Since conversion of the regex matches is done using atoi there's nothing preventing accepting a single-digit hour offset, so be lenient and allow it.

Unit tests pass:

TZ=UTC ninja test
...
OK:        50
FAIL:       0
SKIP:       0
TIMEOUT:    0

Also rebuilt the Fedora 28 tracker package with this patch, it runs fine and does not complain abouta bad date format.

Merge request reports