rapidsai/cudf

[FEA] efficient way to truncate by month

Open

#18,650 建立於 2025年5月3日

在 GitHub 查看
 (5 留言) (0 反應) (1 負責人)C++ (6,000 star) (735 fork)batch import
Pythonfeature requestgood first issuelibcudf

描述

Is there an efficient way to truncate a Series by month / quarter in cuDF?

https://docs.rapids.ai/api/cudf/latest/user_guide/api_docs/api/cudf.core.series.datetimeproperties.floor/ dt.floor doesn't support month / year / quarter

In pandas I can either, depending on the dtype backend:

  • drop to numpy and use astype('datetime64["3M"]'))
  • drop to PyArrow and use pc.floor_temporal

Any suggestion for cuDF?

Example: given

import cudf
from datetime import datetime

s = cudf.Series([datetime(2020, 2, 4)])
s.dt.truncate('3 months')  # Not valid syntax, I know :)

I'd like to end up with

'2020-01-01'

貢獻者指南

技術棧
pythoncpp
領域
dataperformance
議題類型
feature
難度面向新貢獻者的預計實作難度，1 表示很小改動，5 表示專家級工作。
3
預計時間有經驗貢獻者完成調查、實作、測試並準備 pull request 的粗略時間範圍。
half day
活動狀態議題目前的可參與程度：新鮮、活躍、陳舊、阻塞或等待維護者輸入。
blocked
清晰度議題是否清楚說明預期改動、驗收標準和下一步。
clear
前置要求
Basic PythonBasic cuDF
新手友善度1-100 的估計分數，表示該議題對首次貢獻者的友善程度。
40
研究方向
Investigate the current implementation of truncation in cuDF's datetime functionality. Look at the `dt.floor` method and how it currently supports only certain frequencies. The issue suggests a need for month/quarter truncation. Review the relevant C++ source files in `cpp/src/datetime/` and Python wrappers in `python/cudf/cudf/core/series/datetime.py`. Consider if a new kernel is needed or if existing kernels can be extended. Also check if any assignee has started work or if there are discussions in the comments that clarify the approach. The user mentions pandas and PyArrow alternatives, so compare those implementations.
[FEA] efficient way to truncate by month · rapidsai/cudf#18650 | Good First Issue