ARROW-605: [C++] Refactor IPC adapter code into generic ArrayLoader class. Add Date32Type #365
Conversation
Change-Id: I75eb30bdc12723e29f0c8e2c7c92c296dfbe95a7
| explicit TimestampType(TimeUnit unit = TimeUnit::MILLI) | ||
| : FixedWidthType(Type::TIMESTAMP), unit(unit) {} | ||
|
|
||
| explicit TimestampType(const std::string& timezone, TimeUnit unit = TimeUnit::MILLI) |
There was a problem hiding this comment.
Shouldn't timezone default to "UTC"? Also maybe it should be the 2nd parameter?
There was a problem hiding this comment.
We haven't resolved any questions about storing time zones in the IPC metadata (https://github.com/apache/arrow/blob/master/format/Message.fbs), so I did the bare minimum here to support the pandas and R use cases, where they have the notion of tz-naive and tz-aware timestamps. So I don't want to invest any more energy in it until we have a broader discussion about time zones
There was a problem hiding this comment.
Empty string thus indicates a tz-naive timestamp?
There was a problem hiding this comment.
Right, wasn't sure what else to use for a "null" value at the moment
|
fixing compiler warnings |
Change-Id: I6dc42fd7b21d910dd9a8444a02048e4f08ec86cf
Change-Id: I77db11a9c324d4b9055db7490fd3947e651a7d85
|
green build: https://travis-ci.org/wesm/arrow/builds/209347029 |
xhochy
left a comment
xhochy
left a comment
There was a problem hiding this comment.
Except the two questions, this looks good. If they both can be answered with yes, feel free to merge.
| DATE, | ||
|
|
||
| // int32_t days since the UNIX epoch | ||
| DATE32, |
There was a problem hiding this comment.
Is the ordering of the types important yet, i.e. do we match the enums anywhere?
There was a problem hiding this comment.
I don't think so, and I don't think we should guarantee ABI stability of the enum values yet anyhow (at some point we will want to)
| explicit TimestampType(TimeUnit unit = TimeUnit::MILLI) | ||
| : FixedWidthType(Type::TIMESTAMP), unit(unit) {} | ||
|
|
||
| explicit TimestampType(const std::string& timezone, TimeUnit unit = TimeUnit::MILLI) |
There was a problem hiding this comment.
Empty string thus indicates a tz-naive timestamp?
|
thanks |
Author: Korn, Uwe <Uwe.Korn@blue-yonder.com> Closes apache#365 from xhochy/PARQUET-1040 and squashes the following commits: ef359ef [Korn, Uwe] PARQUET-1040: Add missing writer methods Change-Id: I0d7a5b227b64e85c42106e37bb902cc3bbb15e85
3 participants
These are various changes introduced to support the Feather merge in ARROW-452 #361