umputun / rlb-stats / 28690415340

Builds Branch Commit Type Ran Committer Via Coverage
28690415340 fix/candle-precedence fix: clearer error message for out-of-range max_points per review: the client-facing message said "can't parse 'max_points' field" even when the value parsed fine but was out of range (0, 256, negatives). return "max_points must be 1..255" so par... Pull #60 paskal github
83.55
28690346639 fix/aggregate-single-pass perf: drop redundant sort in aggregateCandles emit per review: the input is sorted by StartMinute, so bucket times (floor division of a non-decreasing value) are appended in non-decreasing order and 'order' is already sorted. remove the extra sor... Pull #58 paskal github
84.67
28690266760 fix/aggregator-race test: add dedupe-key collision regression test per review: guards the struct dedup key against a regression to a naive concatenated string key, which would collapse (file=a-b,ip=c) and (file=a,ip=b-c) into one count. Pull #57 paskal github
83.9
28690189759 dashboard-rewrite fix: dispose ECharts instances before HTMX swaps hx-swap="innerHTML" removes the old chart containers on period switch, but their ECharts instances were never disposed, leaking across swaps. dispose any instances inside the swap target in an htmx... Pull #55 paskal github
86.24
28688792735 dashboard-rewrite refactor: harden dashboard rendering per review - fail loudly if the embedded static subtree is missing instead of ignoring the fs.Sub error and serving from a nil FS - render templates into a buffer and return 500 on error, so a template fai... Pull #55 paskal github
86.24
28688682222 fix/bolt-save-merge fix: overwrite unreadable candle on save instead of failing per review: if the existing entry for a key can't be unmarshalled, log a warning and overwrite it rather than returning the error, which would also drop the new candle and prevent recove... Pull #59 paskal github
84.56
28688554483 fix/aggregate-single-pass perf: check cancellation at bucket boundaries in aggregateCandles move the ctx.Done() check into the new-bucket branch so a cancelled request returns only complete buckets, never a partially-filled one. sort a copy of the input by time first so t... Pull #58 paskal github
84.67
28688290763 fix/aggregator-race test: tighten concurrent aggregator test per review correct the comment to say only Store is concurrent (Flush runs after the wait), and assert Flush actually emits a non-empty candle instead of discarding its return. Pull #57 paskal github
83.9
28688182817 fix/candle-precedence test: cover valid max_points path in getCandle add a valid max_points=5 case so the division branch (previously only reached by invalid or aggregate-overridden inputs) is exercised, restoring coverage. Pull #60 paskal github
83.55
28682254032 fix/reject-reserved-dest fix: reject reserved dest "all" on /api/insert "all" is the synthetic node the aggregator writes every record into. POST /api/insert accepted any non-empty dest, so a client posting dest="all" double-counted into that bucket and corrupted the agg... Pull #61 paskal github
83.64
28682097194 fix/aggregate-single-pass test: cover aggregateCandles cancellation and empty-node skip exercises the emit closure's context-cancellation path and empty-bucket skip branch added in the single-pass rewrite, restoring coverage. Pull #58 paskal github
83.97
28682099047 fix/aggregate-single-pass test: cover aggregateCandles cancellation and empty-node skip exercises the emit closure's context-cancellation path and empty-bucket skip branch added in the single-pass rewrite, restoring coverage. Pull #58 paskal github pending completion
28681985834 fix/candle-precedence fix: correct /api/candle aggregate/max_points precedence The README states aggregate overrides max_points, but the handler parsed max_points last so it actually won. Restructure so an explicit aggregate is used and max_points is skipped entirely ... Pull #60 paskal github
83.29
28681705165 fix/bolt-save-merge fix: merge same-minute candles on save instead of overwriting Bolt.Save keyed candles by StartMinute.Unix() and did a plain Put, so a second candle for the same minute overwrote the first. This happens when the aggregator flushes a partial minute... Pull #59 paskal github
83.95
28681490746 fix/aggregate-single-pass perf: aggregate candles in a single pass aggregateCandles rebuilt each output bucket by rescanning the full candle slice, making it O(buckets * candles) - a 30d dashboard (~43k candles, ~100 buckets) rescanned millions of times per render. Bucket... Pull #58 paskal github
83.46