[Enesuygurs]

Description

The graph export averages each series down to the chart width with time_series_subsample(points, self.width). The helper cut the data into slices of round(len(data) / sampling) points and took sampling of them. That only covers the whole series when the division is exact:

• When the rounding goes up, the last slices are empty. t[...][0] raises IndexError (and mean([]) raises StatisticsError in subsample()).
• When it goes down, the slices stop before the end of the series, so the newest points are left out.

With the defaults ([global] history_size=1200, [graph] width=800), both happen. From 801 to 1199 points, the charts end at the 800th sample. Once the history is full, round(1200 / 800) == 2 and slice 600 is data[1200:1202]. From then on, every export cycle logs:

Export graph failed: list index out of range

and no SVG is written again, for any plugin. The TUI's g key fails the same way.

>>> time_series_subsample([(i, float(i)) for i in range(1200)], 800)
IndexError: list index out of range
>>> time_series_subsample([(i, float(i)) for i in range(1000)], 800)[-1]
(799, 799.0)

The slice bounds are now s * len(data) // sampling. Every slice holds at least one point, and together the slices cover the whole series. time_series_subsample() still keeps the first timestamp and the mean value of each slice, and subsample() gets the same change.

>>> time_series_subsample([(i, float(i)) for i in range(1000)], 800)[-1]
(998, 998.5)

Testing:

• Added a new test_017_time_series_subsample and two more checks in test_016_subsample (tests/test_core.py). They fail twice on develop; with the fix, all 8 tests in the file pass.
• A graph export over a full 1200-point history now writes its SVG instead of raising.
• The graph export tests and ruff pass.

Resume

• Bug fix: yes
• New feature: no
• Fixed tickets: none

[nicolargo]

Thanks @Enesuygurs

Merged into develop.