fix: ensure first datapoint is always included in group_by_dynamic #15312
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
MarcoGorelli
force-pushed
the
include-earliest
branch
8 times, most recently
from
91cecb3 to
f81f51a
Compare
MarcoGorelli
commented
Mar 28, 2024
Comment on lines
+13
to
+23
| /// For example, if we have: | ||
| /// | ||
| /// - first datapoint is `2020-01-01 01:00` | ||
| /// - `every` is `'1d'` | ||
| /// - `period` is `'2d'` | ||
| /// - `offset` is `'6h'` | ||
| /// | ||
| /// then truncating the earliest datapoint by `every` and adding `offset` results | ||
| /// in the window `[2020-01-01 06:00, 2020-01-03 06:00)`. To give the earliest datapoint | ||
| /// a chance of being included, we then shift the window back by `every` to | ||
| /// `[2019-12-31 06:00, 2020-01-02 06:00)`. |
There was a problem hiding this comment.
Demo:
In [4]: df = pl.DataFrame({'t': [datetime(2020, 1, 1, 1)], 'i': [0]})
In [5]: df.group_by_dynamic('t', every='1d', period='2d', offset='6h', include_boundaries=True).agg('i')
Out[5]:
shape: (1, 4)
┌─────────────────────┬─────────────────────┬─────────────────────┬───────────┐
│ _lower_boundary ┆ _upper_boundary ┆ t ┆ i │
│ --- ┆ --- ┆ --- ┆ --- │
│ datetime[μs] ┆ datetime[μs] ┆ datetime[μs] ┆ list[i64] │
╞═════════════════════╪═════════════════════╪═════════════════════╪═══════════╡
│ 2019-12-31 06:00:00 ┆ 2020-01-02 06:00:00 ┆ 2019-12-31 06:00:00 ┆ [0] │
└─────────────────────┴─────────────────────┴─────────────────────┴───────────┘Whereas, on the latest release:
In [18]: df.group_by_dynamic('t', every='1d', period='2d', offset='6h', include_boundaries=True).agg('i')
Out[18]:
shape: (0, 4)
┌─────────────────┬─────────────────┬──────────────┬───────────┐
│ _lower_boundary ┆ _upper_boundary ┆ t ┆ i │
│ --- ┆ --- ┆ --- ┆ --- │
│ datetime[μs] ┆ datetime[μs] ┆ datetime[μs] ┆ list[i64] │
╞═════════════════╪═════════════════╪══════════════╪═══════════╡
└─────────────────┴─────────────────┴──────────────┴───────────┘
MarcoGorelli
commented
Mar 28, 2024
Comment on lines
-343
to
+350
| assert_eq!(groups[0], [1, 2]); // 00:00:00 -> 00:30:00 | ||
| assert_eq!(groups[1], [3, 2]); // 01:00:00 -> 01:30:00 | ||
| assert_eq!(groups[2], [5, 2]); // 02:00:00 -> 02:30:00 | ||
| assert_eq!(groups[0], [0, 1]); // (2021-12-15 23:30, 2021-12-16 00:00] | ||
| assert_eq!(groups[1], [1, 2]); // (2021-12-16 00:00, 2021-12-16 00:30] | ||
| assert_eq!(groups[2], [3, 2]); // (2021-12-16 00:30, 2021-12-16 01:00] | ||
| assert_eq!(groups[3], [5, 2]); // (2021-12-16 01:00, 2021-12-16 01:30] |
There was a problem hiding this comment.
here's an example of where the earliest datapoint is currently being excluded
because closed='right', the 00:00:00 point isn't currently included in any windows. This PR fixes that
MarcoGorelli
force-pushed
the
include-earliest
branch
from
9c85e2a to
68c8b94
Compare
MarcoGorelli
commented
Mar 28, 2024
Comment on lines
-167
to
+240
| let (from, to, offset): ( | ||
| let (from, to, offset_fn): ( |
There was a problem hiding this comment.
drive-by: renaming to offset_fn, partially because that what it's called in other places, and also because otherwise it risks being confused with the offset parameter of group_by_dynamic
MarcoGorelli
force-pushed
the
include-earliest
branch
2 times, most recently
from
14471e5 to
b96123b
Compare
MarcoGorelli
commented
Mar 28, 2024
Comment on lines
-151
to
+152
| assert_eq!(groups.len(), 2); | ||
| assert_eq!(groups[1], [2, 2]); | ||
| assert_eq!(groups.len(), 3); | ||
| assert_eq!(groups[1], [1, 1]); |
There was a problem hiding this comment.
same story, earliest datapoint not included in any group due to closed='right'
MarcoGorelli
commented
Mar 28, 2024
Comment on lines
-654
to
+668
| assert_eq!(groups[0], [1, 2]); // 00:00:00 -> 00:30:00 | ||
| assert_eq!(groups[1], [3, 2]); // 01:00:00 -> 01:30:00 | ||
| assert_eq!(groups[2], [5, 2]); // 02:00:00 -> 02:30:00 | ||
| assert_eq!(groups[0], [0, 1]); // (2021-12-15 23:30, 2021-12-16 00:00] | ||
| assert_eq!(groups[1], [1, 2]); // (2021-12-16 00:00, 2021-12-16 00:30] | ||
| assert_eq!(groups[2], [3, 2]); // (2021-12-16 00:30, 2021-12-16 01:00] | ||
| assert_eq!(groups[3], [5, 2]); // (2021-12-16 01:00, 2021-12-16 01:30] |
MarcoGorelli
force-pushed
the
include-earliest
branch
from
b96123b to
7a693f5
Compare
MarcoGorelli
changed the title
github-actions
bot
added
fix
MarcoGorelli
requested review from
ritchie46,
stinodego,
c-peters,
alexander-beedie,
reswqa and
orlp
as code owners
ritchie46
approved these changes
Mar 28, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
closes #15241
With the example from that issue, the output becomes:
which is their expected output
This doesn't break any existing tested behaviour, but does make group-by-dynamic more user-friendly / expected to users using
offsetPerf impact: there's a little extra computation for finding the first window, but that's only for the first window - after that, the windows just keep getting updated by adding
every(no change)