pandas/row-window-on-uneven-times
| Rule | row-window-on-uneven-times |
|---|---|
| Pack | pandas |
| Severity | warning |
| Fix | none, edit by hand |
| Library versions | every version |
A rolling window given as a number counts rows, even with on= naming a time column, so on days with gaps it reaches further back than you meant. Pass the span as an offset, as in rolling("7D", on="day").
weekly = df.rolling(7, on="day")["sold"].sum()weekly = df.rolling("7D", on="day")["sold"].sum()What the rule reports
Section titled “What the rule reports”The rule reports rolling with an integer window and on=. A window given as a number counts rows, also when on= names a time column. rolling(3) without on= gave the same numbers as with it. The table shows the sum for 2024-01-10 on six days with gaps.
| Call | Sum for 2024-01-10 | Rows in the window |
|---|---|---|
rolling(3, on="day").sum() |
14.0 | three rows that reach back 8 days |
rolling("3D", on="day").sum() |
8.0 | the one row of the last 3 days |
The rule misses a row window on a frame with the times in its index, because xpd can't see what the index holds.
How to fix it
Section titled “How to fix it”xpd has no automatic fix, so make the change by hand.
A time window starts with min_periods=1, where the row window gives missing values until it has 7 rows.
Message
Section titled “Message”xpd prints this message with the issue. A name that starts with $ stands for code from your file.
on=$C names the time column, but a window given as a number counts rows. On days with gaps, rolling(3, on="day") summed three rows that reached back 8 days and gave 14.0, where rolling("3D", on="day") gave 8.0 for the last 3 days. Pass the span as an offset, as in rolling("7D", on=$C).Browse the other rules of the pandas pack.