[
https://issues.apache.org/jira/browse/ARROW-270?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15435599#comment-15435599
]
Wes McKinney commented on ARROW-270:
------------------------------------
Some other systems define an absolute "timedelta" type consisting of a
particular number of days, seconds, milliseconds, microseconds, etc. The unit
is fixed, and the timedelta is stored in int64 format
{code}
In [9]: import pandas as pd
In [10]: ts = pd.Timedelta(1000, unit='s')
In [11]: ts
Out[11]: Timedelta('0 days 00:16:40')
In [12]: ts.seconds
Out[12]: 1000
In [13]: ts.asm8 # Internal representation
Out[13]: numpy.timedelta64(1000000000000,'ns')
{code}
What do you think about this kind of data (it would share the same absolute
time units as timestamp, basically)?
> [Format] Define more generic Interval logical type
> --------------------------------------------------
>
> Key: ARROW-270
> URL: https://issues.apache.org/jira/browse/ARROW-270
> Project: Apache Arrow
> Issue Type: Improvement
> Reporter: Wes McKinney
>
> Per discussion in
> https://github.com/apache/arrow/commit/e7e399db5fc6913e67426514279f81766a0778d2#commitcomment-18711366,
> we can create an {{Interval}} type with a unit to be more general.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
