Open Scripter17 opened 3 years ago
A further difficulty with this is that in several cases, parts of the JSON returned from site APIs get added verbatim to the saved metadata. Otherwise some extractors would need a lot more code just to rename/restructure these (sometimes complex) JSON fragments. For this reason, completely unifying the keyword conventions would be much harder than just renaming keys already found in the code.
Deviantart has published_time for the unix timestamp of a post while twitter just... doesn't
Most sites that include some kind of date-time information provide them as date
, which is a datetime
object that can be formatted with strftime format codes ({date:%Y-%m-%d}
)
Timestamps can be converted to datetime
with d
and then formatted like above ({published_time!d:%Y-%m-%d}
) and datetime
objects can be converted to timestamp with T
({date!T}
)
Timestamps can be converted to datetime
Would it be possible to format something like 8-3-2021
(what Behance use I think)?
They don't seem to allow non-zero-padded months and days.
8-3-2021
A pure maintenance and sorting nightmare right there.
(Split off from #1645 because having 2 things in one issue is dumb)
Different sites have wildly different naming conventions for the keywords in
--list-keywords
. This is only made worse by some keywords being ambiguous (such asauthor
anduser
only being different in retweets and stuff)Issues I've found so far:
author[name]
= Reddit'sauthor
= Deviantart'sauthor[username]
published_time
for the unix timestamp of a post while twitter just... doesn'tThis will completely shatter backwards-compatibility with existing databases, but it's something that needs to get addressed eventually and a hypothetical 2.X seems like a good opportunity