Martin Nowak created ARROW-12031:
------------------------------------
Summary: [C++] infer CSV timestamps columns with fractional seconds
Key: ARROW-12031
URL: https://issues.apache.org/jira/browse/ARROW-12031
Project: Apache Arrow
Issue Type: Improvement
Components: C++
Affects Versions: 3.0.0
Reporter: Martin Nowak
After implementing fractional second parsing in and ARROW-9107 and ARROW-10337,
no support for inference of subsecond timestamps was added (see
[code]([https://github.com/apache/arrow/blob/master/cpp/src/arrow/csv/inference_internal.h#L115),
also mentioned in ARROW-3738).
The [InferState logic |
https://github.com/apache/arrow/blob/81d6724546559fa8d6e85376a1cc339dab0506ad/cpp/src/arrow/csv/inference_internal.h#L63]
could test for TimestampNS (with nanosecond resolution) after Timestamps with
second resolution to not change existing inference results.
Wouldn't expect too much interest in trying to infer timestamps with milli- and
microsecond resolution. The nanosecond resolution [can already parse
|https://github.com/apache/arrow/blob/81d6724546559fa8d6e85376a1cc339dab0506ad/cpp/src/arrow/util/value_parsing.h#L512-L518]
the lower resolution ones.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
