[SandroMaglione]

Summary

Breaking change across all four packages. Each transition now holds everything needed to read and edit it.

• branches moves from make groups into a list next to resolve. select.target(path, ...), select.update, select.history, select.initialize, select.stay, and select.decline replace select.<key>, and the compiler checks the calls against the list in both directions. Branch keys, declinable, branch title, and the decline context function are gone.
• { stay: true } and { reenter: true } replace none. { none: true, resolve } becomes { stay: true, commands }.
• commands works on every plain transition.
• update: { owner: value } is a record with one owner, so data only ever means the target's value. A branches entry names the owner as a string.
• guard and decline exist only on on and always, in types and at runtime. A declined invocation outcome used to strand the machine.
• A region's stay preempting a parallel ancestor's handler is existing behaviour, now documented and covered by a public test and a differential case.

Errors name the fix, for example resolve selects undeclared { stay: true }. Add it to branches. The exactness check also rejects undefined, Promise, any, unknown, duplicate entries, and widened paths.

Choices made while building this:

• Retained branchKey and definition key stay, derived from the entry as written (target:Paid, update:root, stay). Coverage and MachineTest.check therefore name branches without an authored key.
• Two old keys that reached one destination are now one entry, so coverage can no longer tell apart two conditions with the same destination.
• no-conditional-none-branch is ported to no-conditional-stay-branch rather than retired, because it still catches a real bug: a disabled control that Machine.can reports as enabled. It flagged a devtools example while I migrated it. no-redundant-resolve is retired.

Type-performance tradeoff: most scenarios are flat or slightly lower. declared transition (10 named branches) rises about 11% and parallel/history/choice about 12%, which pays for checking resolve against branches and owners by path. This is deliberate, per AGENTS.md.

Changeset

• Added or updated for a library or package-metadata change
• Not required because this PR does not change a publishable package

Validation

• pnpm check
• Automated type-performance measurement passed or was not required
• Automated runtime- and memory-performance measurement passed or was not required

🤖 Generated with Claude Code

[github-actions]

Type performance

Measured with TypeScript 6.0.3 and skipLibCheck=true.

Scenario	Base	PR	Difference
Effect only	55	55	0 (0.0%)
Import effect-machine	55	55	0 (0.0%)
Machine.state (3 children)	2,664	2,664	0 (0.0%)
Machine.make (3 states, 2 events)	8,932	8,849	-83 (-0.9%)
machine.handle (3 states, 2 transitions)	39,449	38,693	-756 (-1.9%)
declared transition (10 named branches)	60,402	67,072	+6,670 (+11.0%)
atomic construction, guards, and reentry	85,710	77,731	-7,979 (-9.3%)
registered invocation (input-taking Effect)	39,819	38,733	-1,086 (-2.7%)
machine.handle (depth 24)	194,531	194,030	-501 (-0.3%)
machine.handle (wide depth 16)	218,238	217,737	-501 (-0.2%)
machine.handle (parallel/history/choice)	91,642	102,455	+10,813 (+11.8%)
machine definition (3 independent implementations)	99,173	92,828	-6,345 (-6.4%)
machine exact input/output/error/services	82,941	80,179	-2,762 (-3.3%)
execution adapter readiness	71,428	73,484	+2,056 (+2.9%)

Marginal instantiations are measured against the matching setup without that API call:

Scenario	Base	PR	Difference
Import effect-machine	0	0	0
Machine.state (3 children)	2,609	2,609	0 (0.0%)
Machine.make (3 states, 2 events)	6,260	6,177	-83 (-1.3%)
machine.handle (3 states, 2 transitions)	30,517	29,844	-673 (-2.2%)
declared transition (10 named branches)	51,849	58,602	+6,753 (+13.0%)
atomic construction, guards, and reentry	70,370	62,474	-7,896 (-11.2%)
registered invocation (input-taking Effect)	32,682	31,679	-1,003 (-3.1%)
machine.handle (depth 24)	177,222	176,804	-418 (-0.2%)
machine.handle (wide depth 16)	201,707	201,289	-418 (-0.2%)
machine.handle (parallel/history/choice)	79,144	90,040	+10,896 (+13.8%)
machine definition (3 independent implementations)	88,492	82,230	-6,262 (-7.1%)
machine exact input/output/error/services	70,732	68,053	-2,679 (-3.8%)
execution adapter readiness	31,922	31,884	-38 (-0.1%)

Check times (informational)

Scenario	Base	PR
Effect only	0.02s	0.02s
Import effect-machine	0.02s	0.02s
Machine.state (3 children)	0.06s	0.07s
Machine.make (3 states, 2 events)	0.12s	0.12s
machine.handle (3 states, 2 transitions)	0.22s	0.22s
declared transition (10 named branches)	0.29s	0.33s
atomic construction, guards, and reentry	0.38s	0.37s
registered invocation (input-taking Effect)	0.23s	0.22s
machine.handle (depth 24)	0.47s	0.46s
machine.handle (wide depth 16)	0.50s	0.52s
machine.handle (parallel/history/choice)	0.38s	0.41s
machine definition (3 independent implementations)	0.40s	0.38s
machine exact input/output/error/services	0.36s	0.38s
execution adapter readiness	0.32s	0.33s

Type instantiations are the comparison metric. Check time varies with runner load and is informational only.

[github-actions]

Runtime performance

Median of 5 independent benchmark processes on AMD EPYC 7763 64-Core Processor with Node v24.22.2.

Pull request baseline

Scenario	Effect Machine
Plan counter transitions	114,650 transitions/s
Drain burst with terminal fence	398,207 increments/s
Drain burst with a change observer	368,815 increments/s
Lookup and send to one child	290,535 increments/s
Start and stop a machine	117,426 machines/s
Start and stop a parent with one child	23,765 families/s
Complete an Effect invocation with tracing enabled	6,302 invocations/s
Complete an Effect invocation with tracing disabled	7,470 invocations/s
Plan transitions through a compound state	99,921 transitions/s
Plan transitions through parallel regions	79,062 transitions/s
Drain burst through a compound state	339,256 events/s
Drain burst through two parallel regions	312,941 events/s
Drain a compound-state burst with a change observer	314,273 events/s

Process runtime reference points

Scenario	Effect Machine
Start and stop a raw generic process	30,055 processes/s
Start and stop a raw compiled process	69,075 processes/s

Memory profile	Effect Machine
Idle machine	1.8 KiB
Raw generic managed process	11.4 KiB
Raw compiled process	3.1 KiB
Two independent idle machines	3.6 KiB
Idle parent with one child	6.0 KiB
Parent with observed child registry	9.6 KiB
Parent with observed invoked child snapshots	7.0 KiB

Effect Machine change from base

Metric	Base	Base variability	PR	PR variability	Difference
Plan counter transitions	113,379 transitions/s	1.2% MAD	114,650 transitions/s	1.0% MAD	+1.1%
Drain burst with terminal fence	403,670 increments/s	0.3% MAD	398,207 increments/s	0.8% MAD	-1.4%
Drain burst with a change observer	371,080 increments/s	0.2% MAD	368,815 increments/s	0.7% MAD	-0.6%
Lookup and send to one child	292,313 increments/s	0.2% MAD	290,535 increments/s	1.0% MAD	-0.6%
Start and stop a machine	116,877 machines/s	0.8% MAD	117,426 machines/s	0.5% MAD	+0.5%
Start and stop a parent with one child	24,116 families/s	1.9% MAD	23,765 families/s	0.3% MAD	-1.5%
Complete an Effect invocation with tracing enabled	6,316 invocations/s	0.9% MAD	6,302 invocations/s	1.1% MAD	-0.2%
Complete an Effect invocation with tracing disabled	7,494 invocations/s	0.9% MAD	7,470 invocations/s	0.5% MAD	-0.3%
Plan transitions through a compound state	100,194 transitions/s	0.7% MAD	99,921 transitions/s	1.7% MAD	-0.3%
Plan transitions through parallel regions	78,140 transitions/s	0.2% MAD	79,062 transitions/s	1.1% MAD	+1.2%
Drain burst through a compound state	344,197 events/s	0.1% MAD	339,256 events/s	0.6% MAD	-1.4%
Drain burst through two parallel regions	316,332 events/s	0.4% MAD	312,941 events/s	0.6% MAD	-1.1%
Drain a compound-state burst with a change observer	316,638 events/s	0.4% MAD	314,273 events/s	0.4% MAD	-0.7%
Idle machine heap per unit	1.8 KiB	0.0% MAD	1.8 KiB	0.2% MAD	+0.2%
Raw generic managed process heap per unit	11.3 KiB	0.0% MAD	11.4 KiB	0.0% MAD	+0.9%
Raw compiled process heap per unit	3.1 KiB	0.0% MAD	3.1 KiB	0.1% MAD	+0.0%
Two independent idle machines heap per unit	3.6 KiB	1.1% MAD	3.6 KiB	0.2% MAD	-0.0%
Idle parent with one child heap per unit	6.0 KiB	0.0% MAD	6.0 KiB	0.0% MAD	+0.0%
Parent with observed child registry heap per unit	9.6 KiB	0.0% MAD	9.6 KiB	0.0% MAD	+0.0%
Parent with observed invoked child snapshots heap per unit	7.0 KiB	0.0% MAD	7.0 KiB	0.1% MAD	-0.2%

Process runtime reference change from base

Metric	Base	Base variability	PR	PR variability	Difference
Start and stop a raw generic process	30,495 processes/s	1.0% MAD	30,055 processes/s	1.4% MAD	-1.4%
Start and stop a raw compiled process	67,395 processes/s	0.3% MAD	69,075 processes/s	0.2% MAD	+2.5%

Regression guard

No large, noise-adjusted throughput or heap regressions detected.

Versions and interpretation

• Effect Machine: 0.46.0

Higher throughput is better; lower heap is better. Variability is the median absolute deviation across independent processes, relative to their median. Small differences on shared GitHub-hosted hardware remain informational; the required guard rejects only large changes beyond the measured noise allowance.