apache / apache/arrow-java

[Java] getObject API for temporal types is inconsistent and in some cases incorrect

オープン
#324 コメント 3 件 リアクション 0 件 担当者 0 名 GitHub で見る
Type: bug
主要言語
Java
スター
94
フォーク
152
平均マージ
3日 16時間
マージ済み PR(30日)
11

説明

It seems that the work which has been tracked in ARROW-2015 and merged in to change the return types of the various Time and Date vector types when using the getObject API missed some of the vector types which are temporal and so should return a temporal type, and provided an incorrect implementation for others (some of this was pointed out in the initial PR review, but it seems that it slipped through the cracks and was not addressed before merging).

Here is a table of the various temporal vector types, what they currently return from getObject, and what they should return, in my opinion (I have included ones in which the implementation is correct for completeness, and coloured them green).

 

 

|Vector class|Current return type|Proposed return type|Comments|
|
|-|-|-|-|-|
|DateDayVector|Integer|LocalDate|Currently returns the raw value of days since epoch, should return the actual date|
|
|DateMilliVector|LocalDateTime|LocalDate|This type is supposed to encode a date, not a datetime, so even though epoch millis are used, the result should be a LocalDate|
|
|DurationVector|Duration|Duration|Correct.|
|
|IntervalDayVector|Duration|Period|As per , Interval should be a calendar-based datatype, not a time-based one. This is represented in Java by a Period type. However, I note that the Java Period class does not support milliseconds, unlike the Arrow type, which might be why Duration is being returned. Some discussion may be needed on the best way to deal with this.|
|
|IntervalYearVector|Period|Period|Correct.|
|
|TimeMicroVector|Long|LocalTime|Currently returns the raw number of micros, should return the actual time|
|
|TimeMilliVector|LocalDateTime|LocalTime|Currently returns a datetime on 1970-01-01 with the correct time component, should just return the time|
|
|TimeNanoVector|Long|LocalTime|Currently returns the raw number of nanos, should return the actual time|
|
|TimeSecVector|Integer|LocalTime|Currently returns the raw number of seconds, should return the actual time|
|
|TimeStampMicroVector|LocalDateTime|LocalDateTime|Correct.|
|
|TimeStampMilliVector|LocalDateTime|LocalDateTime|Correct.|
|
|TimeStampNanoVector|LocalDateTime|LocalDateTime|Correct.|
|
|TimeStampSecVector|LocalDateTime|LocalDateTime|Correct.|
|
|TimeStampMicroTZVector|Long|ZonedDateTime|Currently returns the underlying micros, and TZ has to be obtained separately. Should return the actual datetime with timezone|
|
|TimeStampMilliTZVector|Long|ZonedDateTime|Currently returns the underlying millis, and TZ has to be obtained separately. Should return the actual datetime with timezone|
|
|TimeStampNanoTZVector|Long|ZonedDateTime|Currently returns the underlying nanos, and TZ has to be obtained separately. Should return the actual datetime with timezone|
|
|TimeStampSecTZVector|Long|ZonedDateTime|Currently returns the underlying seconds, and TZ has to be obtained separately. Should return the actual datetime with timezone|



I am happy to submit a PR to fix this if there is no other reason for this to persist other than these types being rarely used, but note that these would all be breaking API changes.



 |

**Reporter**: [Matt Jadczak](https://issues.apache.org/jira/browse/ARROW-9915)

**Note**: *This issue was originally created as [ARROW-9915](https://issues.apache.org/jira/browse/ARROW-9915). Please see the [migration documentation](https://github.com/apache/arrow/issues/14542) for further details.*

コントリビューションガイド

コントリビューションガイドを開く

調査の方向性

まず、一覧にある時間ベクトルクラスの getObject 実装を見つけ、ARROW-2015 と pull request 2966 による変更を確認します。Java の戻り値の型を決める前に、Arrow の interval スキーマ、特に IntervalDayVector のミリ秒表現を確認してください。変更対象のベクトルが提案された時間型を一貫して返し、変更された各マッピングに対するカバレッジがあれば完了です。

索引モデルが issue の本文から書いたものです。

評価

技術スタック
java
領域
api
issue の種類
バグ
難易度
4/5
見積もり時間
3〜5日
活発さ
停滞
明瞭さ
おおむね明確
初心者へのやさしさ
35/100

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。