max / makenotwork

Log the discarded tier-price install, and correct two stale docs tier_prices.rs: install_test_default discarded the OnceLock Err, so a racing install was invisible. Warn and keep the first table, matching install_global. Losing the race is the documented outcome, not a fault. tagtree docs: the TagIndex section described a two-tier suggestion strategy the code outgrew. Document the real three tiers and the bulk operations. pom serve.rs: the dead-letter delete reused the post-delivery wording. A failed delete there re-drops the row, it does not re-send.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Author: Max Johnson <me@maxj.phd> · 2026-07-30 23:49 UTC

Signed with PGP, not checked

Commit: 2b730c53d989d3f96124396c8226ca950c9d285a

Parent: 8104661

3 files changed, +27 insertions, -8 deletions

M server/src/tier_prices.rs +10 -5

			@@ -150,7 +150,8 @@
150	150		// once, so a second call means two config sources exist and the
151	151		// second one is being ignored.
152	152		tracing::warn!(
153		-	"TierPrices::install_global called more than once; keeping the first install"
	153	+	"TierPrices::install_global called a second time; the new table is ignored and \
	154	+	the first install stays in force"
154	155		);
155	156		}
156	157		}
			@@ -179,10 +180,14 @@
179	180		let a = Assumptions::load("docs/business/assumptions.toml")
180	181		.expect("test setup: load canonical assumptions.toml");
181	182		let tp = TierPrices::from_assumptions(&a);
182		-	// Ignored deliberately: concurrent tests race to install here and the
183		-	// values are identical, so losing the race is the documented outcome
184		-	// rather than a failure.
185		-	let _ = GLOBAL.set(tp);
	183	+	// Losing the race is the documented outcome rather than a failure:
	184	+	// concurrent tests install identical values.
	185	+	if GLOBAL.set(tp).is_err() {
	186	+	tracing::warn!(
	187	+	"TierPrices::install_test_default raced a concurrent install; the fixture is \
	188	+	ignored and the first table stays in force"
	189	+	);
	190	+	}
186	191		}
187	192		}
188	193

M pom/src/cli/serve.rs +1 -1

			@@ -179,7 +179,7 @@
179	179		alert_id = p.id,
180	180		alert_key = %p.alert_key,
181	181		error = %e,
182		-	"pending alert delete failed, alert will re-send"
	182	+	"pending alert delete failed, the drop retries next tick"
183	183		);
184	184		}
185	185		} else {

M shared/tagtree/docs/architecture.md +16 -2

			@@ -74,12 +74,26 @@
74	74
75	75		### TagIndex (Autocomplete)
76	76
77		-	Sorted `Vec<String>` with binary-search lookup. Two-tier suggestion strategy:
	77	+	Sorted `Vec<String>` with binary-search lookup. Three suggestion tiers, each running only if the one before it left room under `limit`:
78	78
79	79		1. Path prefix (tier 1): binary search for tags whose full path starts with input. O(log n + k).
80	80		2. Segment prefix (tier 2): linear scan for tags where any non-first segment starts with input. Only runs if tier 1 didn't fill the limit and input has no dots. A segment index gates entry: if no known segment starts with input, scan is skipped in O(log m).
	81	+	3. Fuzzy segment (tier 3): linear scan for tags with any segment within Levenshtein distance 2 of input, sorted by distance then lexicographically. Skips tags already returned by tiers 1 and 2.
81	82
82		-	Key methods: `new()`, `empty()`, `insert()`, `remove()`, `rebuild()`, `contains()`, `suggest(input, limit)`, `suggest_with_status(input, limit)`.
	83	+	`suggest(input, limit)` runs tiers 1 and 2. `suggest_fuzzy(input, limit)` runs all three: it calls `suggest` first and appends tier-3 hits. Within each tier results are sorted, and earlier tiers come first.
	84	+
	85	+	Key methods: `new()`, `empty()`, `insert()`, `remove()`, `rebuild()`, `contains()`, `len()`, `is_empty()`, `iter()`, `suggest(input, limit)`, `suggest_with_status(input, limit)` (adds a flag for whether input is itself a tag), `suggest_fuzzy(input, limit)`.
	86	+
	87	+	`insert()` and `remove()` keep sorted order. `remove()` does not prune orphaned segments; call `rebuild()` when segment accuracy matters after many removals.
	88	+
	89	+	### Bulk Operations (free functions over `&mut TagIndex`)
	90	+
	91	+	Each rewrites the index in place and returns a count.
	92	+
	93	+	- `rename_prefix_bulk(old, new, index)`: apply `rename_prefix` to every tag, rebuilding the index. Returns tags modified.
	94	+	- `remove_subtree(prefix, index)`: drop `prefix` and all its descendants. Returns tags removed.
	95	+	- `merge_tags(source, target, index)`: rewrite the `source` prefix to `target` across the subtree, deduplicating collisions with tags already under `target`. Implemented as a call through to `rename_prefix_bulk`. Returns tags moved.
	96	+	- `batch_rename(operations, index)`: apply `(old, new)` prefix renames in order. Returns the sum of per-operation counts, so a tag renamed twice counts twice.
83	97
84	98		## Integration Patterns
85	99