FlexMeasures / flexmeasures / 37625671193

87%

push

github

Refactor: Move forecast cycle persistence into the forecasting service (#2645)

* Refactor forecast cycle persistence into the forecasting service

Keep legacy pipeline and plugin entrypoints compatible while exposing computation-only predictions and cycles. Preserve per-cycle commits and use transient source attribution for synchronous dry runs.

Refs #2486. Validated with 94 forecasting tests and pre-commit hooks.

* docs/changelog: link forecast persistence entry to PR #2645

* Keep a forecaster's transient source with its organisation

Since #2634 an organisation is part of what identifies a data source, and
`_persist_data_source_id` now resolves `refresh_data_source(self.forecast_source())`
rather than the `data_source` property, which passed `source_account` along.
Without the account, the lookup asks for a source belonging to nobody, does not
find the forecaster's own, and creates an unowned one beside it: every queued
forecast would record under a source that says it belongs to no organisation,
and a `source-account` filter would stop matching the forecasts it used to.

The test has to own its asset to see any of this: `setup_fresh_test_forecast_data`
builds assets without an organisation, so the existing forecasting tests pass
whether or not the account is carried.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0129WrXeJ5gia2pctFH93BqC
Signed-off-by: F.N. Claessen <claessen@seita.nl>

* Say when a cycle saved no forecast at all

`save_forecast` returned on an empty frame before any logging, so a cycle that
computed nothing was indistinguishable in the log from a cycle that never ran.
The dry-run path already reports an empty forecast, through
`_log_forecast_dry_run`, so the saving path should not be the quiet one.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0129WrXeJ5gia2pctFH93BqC
Signed-off-by: F.N. Claessen <claessen@seita.nl>

... (continued)

105 of 111 new or added lines in 5 files covered. (94.59%)

21170 of 24486 relevant lines covered (86.46%)

0.86 hits per line