jorisvandenbossche commented on code in PR #37822:
URL: https://github.com/apache/arrow/pull/37822#discussion_r1639997184


##########
python/pyarrow/tests/test_compute.py:
##########
@@ -2481,6 +2493,10 @@ def _check_temporal_rounding(ts, values, unit):
 
 @pytest.mark.skipif(sys.platform == "win32" and not util.windows_has_tzdata(),
                     reason="Timezone database is not installed on Windows")
[email protected](
+    sys.platform == "emscripten",
+    reason="Emscripten datetime is implemented in Javascript and works 
differently"
+)

Review Comment:
   Ah, good catch. I should have thought about this, because we have discussed 
related issues in the past. We explicitly don't use the `tzdata` package 
because of the compatibility issue. Although the discussion at 
https://github.com/apache/arrow/issues/31472 / 
https://github.com/apache/arrow/issues/28868#issuecomment-1378032344 is 
specifically for Windows (where getting the tzdata through a package would be 
especially useful, given we now require users to download the text database 
manually), and from re-reading that it seems that the binary format not being 
supported is only a Windows issue, and for linux it should be fine. But 
generally for linux the system provides the tz data, so we don't actually test 
that with the `tzdata` package. I am wondering if we would have similar 
problems as you encountered here for Linux with `tzdata` .. (given the `slim` / 
`fat` difference)
   
   > I'm inclined to bundle pre-built timezones into pyarrow build as resources?
   
   I am not entirely sure this would actually be possible.
   
   For Windows, we do provide a helper function to both set the path to the tz 
database manually to a custom path (`pyarrow.set_timezone_db_path`), and then 
also a helper to download the data (`pyarrow.util.download_tzdata_on_windows`). 
   
   But that requires you to be able to point to a custom path, and right now 
that is only possible on Windows AFAIK
   
   



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to