[marandaneto]

💡 Motivation and Context

Go can return a definitive local flag result that disagrees with the flags service. For example, RE2's ^\d+$ does not match ١, while the service's Unicode-aware regex does. Lookarounds, backreferences and class-local backspace also compile on the service but not in Go. Their compilation failures previously became false, nil, which skipped remote fallback and could become a match through cohort negation.

This adapts the agreed local-subset and remote-fallback approach from JS #5354 and its cross-SDK discussion to Go's RE2 engine. It does not depend on that PR being merged or released.

The regex cache now retains capability metadata alongside compiled patterns. Unsupported syntax, differing set and repetition parsing, repeated assertions, and Unicode or vertical-tab inputs with RE2 shorthand classes use the existing server-required error path. Word boundaries also defer because they can exhaust the service's backtracking budget even on short ASCII inputs. GetFeatureFlag can then request a remote result. Unsupported regex now preserves the existing RequiresServerEvaluationError category through matcher, cohort and dependency boundaries. This prevents an unresolved earlier regex from being replaced by a later catch-all variant. Ordinary inconclusive or missing-property conditions still allow independently matching later rules, as required by the evaluator contract. Local-only unsupported-regex calls remain unresolved, including through negated cohorts.

Repetition candidates are scanned forward without repeatedly searching or splitting the remaining suffix. This bounds the reviewed braces-and-commas case without changing the regex cache. Unterminated braced x/p/P escapes now stop classification at the first missing close and reach the existing compiler categorization once. Bare malformed p/P patterns stay definitively invalid, while unsupported verbose-mode carriers still defer. Proven reversed uint32 repetition bounds such as a{2,1} now fail both operators without a remote request. RE2 size caps, overflow and leading-zero uncertainty still require the server. There is no negative cache or new parser framework. Known invalid shared syntax still fails both regex operators. Negative operators only negate a valid match. Ordinary ASCII patterns, escaped punctuation, supported flags, strict anchors, Unicode property classes and scalar matching remain local. Empty patterns, explicit null matching and existing stringification are preserved. This reduces local coverage conservatively without adding an engine, dependency or public option. It does not claim universal regex, Unicode-table or resource-budget equivalence.

Malformed unescaped brace candidates such as a{x}, a{, a{2 } and a{1,2 } now require a decimal lower bound and a closing } before local compilation is eligible. Valid {n}, {n,m} and {n,} forms remain eligible, as do escaped and class-local literal braces. Contract limitation relative to the local evaluator spec: the spec requests definitive false for invalid regex syntax; these ambiguous brace forms conservatively defer to the service instead, because Go accepts them as literals. Local-only evaluation remains unresolved rather than returning the wrong definitive boolean.

💚 How did you test it?

• Added regressions before production changes and recorded RED failures for wrong local booleans and skipped fallback. Focused compatibility tests now pass for both operators, direct flags, cohorts, negated cohorts and local-only calls. Only the remote request boundary is stubbed. Missing-property and portable-pattern controls verify the fallback gate is not always on or always off.

• Compared both operators against the frozen actual-service oracle across 4,868 expanded vectors, 19 review vectors and 72 additional probes. All definitive results agreed. These corpora overlap and are finite.

• Added ordered two-variant regressions before the feedback fix. They returned second locally for ^a\b$ against a before the change. They now reach remote fallback, or remain server-required in local-only mode, through direct, cohort, nested-cohort and dependency paths. Missing-context, safe-local and invalid-pattern controls remain local as appropriate.

• A standard Go benchmark for 4,000 opening braces followed by 4,000 commas and } allocated about 262 MB/op and took 145-149 ms/op before the scanner fix. It now allocates 0 B/op and 0 allocations/op. Steady measurements were about 6.8-7.0 microseconds/op on this machine, with approximately linear scaling at 1,000/2,000/4,000 braces. Timings are observations, not unit-test thresholds or a universal complexity guarantee.

• Added actual-service-verified reversed-bound regressions before the fix. Both operators previously requested remote evaluation or returned unresolved locally. They now return definitive false for a single rule and permit the later ordered variant, including local-only calls. Large valid counts and native nested-count limits still defer.

• Benchmarked repeated unterminated x/p/P escapes and an unsupported (?x)# comment carrier at 12/48/192 KiB, both classification and actual property matching. At 192 KiB the old classifier medians were about 64-74 ms/op. Steady fixed medians were about 2 microseconds/op for classification, 4-5 microseconds for x/comment matching, and 124-135 microseconds for p/P matching including native parsing. This is bounded, machine-specific evidence rather than a universal regex latency guarantee. Semantic category, class and escape controls pass before and after.

• After merging latest origin/main, ran go test -mod=readonly -count=1 -run '^TestRegexCompatibility' ., full root make build with race tests and vet, make api-diff, gofmt and diff checks. All passed on Go 1.26.4, macOS arm64.

• Completed diff-scoped deslop and test-audit Authoring checks. Final branch autoreview passed on committed HEAD c8126ce3c3c4ad09ffc4bc8930c98efc5cbd17e4 against origin/main at 56aa91d692651b853c9c9857517a9cda077392a8 with no actionable findings.

• Follow-up review fix (680ba36): added regressions before production edits for malformed brace patterns, both operators, remote/local-only evaluation, negated cohorts and ordered variants through nested cohorts and dependencies. They failed with definitive local results and zero remote calls before the fix; the same commands and full compatibility suite pass afterward. Retained escaped/class-local brace and valid repetition controls. Re-ran make build (root race tests, vet and build), make api-diff, gofmt and diff checks on Go 1.26.4/macOS arm64; all passed. The existing repetition-candidate benchmark still reports zero allocations and now stops at the first unsupported brace.

The minimum Go version and other platforms were not run locally. Separate otel and posthogmcpsdk module checks were not run because this changes neither their code nor the core public API. No live-service tests were used. GitHub CI results are separate from these local checks.

📝 Checklist

• All commits are signed (see Commit signing).
• I reviewed the submitted code.
• I added tests to verify the changes.
• I updated the docs if needed. No public API or configuration change requires documentation.
• No breaking change or entry added to the changelog.

If releasing new changes

• Ran pnpm changeset to generate a changeset file. The equivalent posthog-go patch Changeset is included directly. No version was bumped.

🤖 Agent context

Autonomy: Human-driven (agent-assisted)

The operator directed this Go port and its publication. Pi assisted with implementation and test-audit using Git, GitHub CLI and Go tooling. An isolated Pi autoreview checked the final committed branch. The implementation keeps RE2 and uses existing remote fallback rather than adding another engine or translating unsupported patterns. Human review is still required.

[github-actions]

posthog-go Compliance Report

Date: 2026-10-23 10:41:27 UTC
Duration: 169322ms

✅ All Tests Passed!

111/111 tests passed

---

Capture_V1 Tests

✅ 94/94 tests passed

View Details

Test	Status	Duration
Endpoint And Method.Targets V1 Endpoint	✅	19ms
Endpoint And Method.Does Not Use Legacy Endpoints	✅	7ms
Required Headers.Has Authorization Bearer Header	✅	18ms
Required Headers.Has Content Type Json	✅	8ms
Required Headers.Has Posthog Sdk Info Format	✅	6ms
Required Headers.Has Posthog Attempt Header	✅	7ms
Required Headers.Has Posthog Request Id	✅	6ms
Required Headers.Has Posthog Request Timestamp	✅	7ms
Required Headers.Has User Agent	✅	7ms
Body Format.Body Has Created At And Batch	✅	6ms
Body Format.No Api Key In Body	✅	7ms
Body Format.No Sent At In Body	✅	8ms
Event Format.Event Has Required Root Fields	✅	16ms
Event Format.Event Uuid Is Valid	✅	7ms
Event Format.Event Timestamp Is Rfc3339	✅	7ms
Event Format.Distinct Id Is String	✅	7ms
Event Format.Distinct Id At Root Not Properties	✅	6ms
Event Format.Custom Properties Preserved	✅	7ms
Event Format.Set Properties Preserved	✅	7ms
Event Format.Set Once Properties Preserved	✅	7ms
Event Format.Groups Properties Preserved	✅	7ms
Event Format.Sdk Generates Uuid If Not Provided	✅	7ms
Event Format.Event Has Required Root Fields Batch	✅	9ms
Event Format.Event Uuid Is Valid Batch	✅	10ms
Event Format.Event Timestamp Is Rfc3339 Batch	✅	9ms
Event Format.Distinct Id Is String Batch	✅	9ms
Event Format.Distinct Id At Root Not Properties Batch	✅	10ms
Event Format.Custom Properties Preserved Batch	✅	9ms
Event Format.Set Properties Preserved Batch	✅	10ms
Event Format.Set Once Properties Preserved Batch	✅	10ms
Event Format.Groups Properties Preserved Batch	✅	9ms
Event Format.Sdk Generates Uuid If Not Provided Batch	✅	9ms
Batch Behavior.Multiple Events In Single Batch	✅	32ms
Batch Behavior.Batch Envelope Smoke	✅	10ms
Batch Behavior.Flush With No Events Sends Nothing	✅	4ms
Batch Behavior.Flush At Triggers Batch	✅	1008ms
Batch Behavior.Created At Reflects Batch Creation Time	✅	18ms
Deduplication.Generates Unique Uuids	✅	17ms
Deduplication.Different Events Same Content Different Uuids	✅	9ms
Deduplication.Preserves Uuid On Retry	✅	5135ms
Deduplication.Preserves Timestamp On Retry	✅	5134ms
Deduplication.Preserves Uuid And Timestamp On Batch Retry	✅	5139ms
Deduplication.No Duplicate Events In Batch	✅	34ms
Header Behavior On Retry.Attempt Header Starts At One	✅	19ms
Header Behavior On Retry.Attempt Header Increments On Retry	✅	10137ms
Header Behavior On Retry.Request Id Preserved On Retry	✅	5137ms
Header Behavior On Retry.Different Requests Have Different Request Ids	✅	2018ms
Header Behavior On Retry.Request Timestamp Changes On Retry	✅	5134ms
Response Format Validation.Success Response Has Uuid Keyed Results	✅	8ms
Response Format Validation.Success Response Has Ok For Each Event	✅	29ms
Response Format Validation.Success No Retry After When All Ok	✅	28ms
Response Format Validation.Success Retry After Present When Retry Events	✅	131ms
Response Format Validation.Success No Retry After When Drop Only	✅	29ms
Response Format Validation.Response Echoes Request Id	✅	9ms
Retry Behavior.Retries On 408	✅	5136ms
Retry Behavior.Retries On 500	✅	5138ms
Retry Behavior.Retries On 503	✅	5137ms
Retry Behavior.Retries On 504	✅	5137ms
Retry Behavior.Retryable Errors Have Retry After	✅	2134ms
Retry Behavior.Respects Retry After On Retryable Error	✅	8141ms
Retry Behavior.Does Not Retry On 400	✅	2133ms
Retry Behavior.Does Not Retry On 401	✅	2135ms
Retry Behavior.Does Not Retry On 402	✅	2134ms
Retry Behavior.Does Not Retry On 413	✅	2134ms
Retry Behavior.Does Not Retry On 415	✅	2134ms
Retry Behavior.Non Retryable Errors Have No Retry After	✅	2134ms
Retry Behavior.Implements Backoff	✅	15134ms
Retry Behavior.Max Retries Respected	✅	15137ms
Partial Batch Handling.Handles 200 Full Success	✅	2011ms
Partial Batch Handling.Handles 200 With All Ok	✅	3034ms
Partial Batch Handling.Does Not Retry Dropped Events	✅	3029ms
Partial Batch Handling.Does Not Retry Limited Events	✅	3030ms
Partial Batch Handling.Prunes Ok Events On Partial Retry	✅	5138ms
Partial Batch Handling.Prunes Dropped Events On Partial Retry	✅	5137ms
Partial Batch Handling.Retries Only Retry Events From Partial	✅	5139ms
Partial Batch Handling.Partial Retry Preserves Uuids	✅	5137ms
Partial Batch Handling.Partial Retry Attempt Header Increments	✅	5135ms
Partial Batch Handling.Partial Retry Request Id Preserved	✅	5135ms
Partial Batch Handling.Respects Retry After On Partial	✅	5135ms
Partial Batch Handling.Unknown Result Treated As Terminal	✅	3030ms
Partial Batch Handling.Mixed Ok Drop Limited No Retry	✅	3034ms
Compression.Sends Gzip Content Encoding	✅	17ms
Compression.No Content Encoding When Disabled	✅	6ms
Compression.Compressed Body Is Decompressible	✅	16ms
Error Handling.Does Not Retry On Unknown 4Xx	✅	2131ms
Event Options.Cookieless Mode Override	✅	8ms
Event Options.Disable Skew Correction Override	✅	6ms
Event Options.Process Person Profile Override	✅	5ms
Event Options.Product Tour Id Override	✅	6ms
Event Options.Unset Options Omitted	✅	6ms
Event Options.Options Override In Batch	✅	9ms
Geoip And Historical Migration.Geoip Disable Injected Into Properties	✅	6ms
Geoip And Historical Migration.Historical Migration Set In Body	✅	6ms
Geoip And Historical Migration.Historical Migration Absent By Default	✅	5ms

Feature_Flags Tests

✅ 17/17 tests passed

View Details

Test	Status	Duration
Request Payload.Request With Person Properties Device Id	✅	16ms
Request Payload.Flags Request Uses V2 Query Param	✅	16ms
Request Payload.Flags Request Hits Flags Path Not Decide	✅	16ms
Request Payload.Flags Request Omits Authorization Header	✅	16ms
Request Payload.Token In Flags Body Matches Init	✅	15ms
Request Payload.Groups Round Trip	✅	16ms
Request Payload.Groups Default To Empty Object	✅	16ms
Request Payload.Disable Geoip False Propagates As Geoip Disable False	✅	16ms
Request Payload.Disable Geoip Omitted Defaults To False	✅	16ms
Request Payload.Flag Keys To Evaluate Contains Only Requested Key	✅	16ms
Request Lifecycle.No Flags Request On Init Alone	✅	2ms
Request Lifecycle.No Flags Request On Normal Capture	✅	7ms
Request Lifecycle.Two Flag Calls Produce Two Remote Requests	✅	29ms
Request Lifecycle.Mock Response Value Is Returned To Caller	✅	15ms
Retry Behavior.Retries Flags On 502	✅	18ms
Retry Behavior.Retries Flags On 504	✅	17ms
Side Effect Events.Get Feature Flag Captures Feature Flag Called Event	✅	17ms

[posthog]

PostHog Review · Standard · reviewed 0e4b57d: 2 Should fix.

This is the only automatic review while this pull request is a draft. Comment @posthog review to review new changes sooner. Otherwise, a review of the new changes runs when you mark it ready for review.

[posthog]

PostHog Review · Standard · reviewed b26b843: 2 Should fix.

This Standard review started from the posthog-review label. Comment @posthog review --deep for a Deep review, or @posthog help to see all commands.

[veria-ai]

PR overview

All previously flagged issues have been addressed. No open security concerns remain on this pull request.

Security review

No open security issues remain on this pull request.

Fixed/addressed: 1 · PR risk: 0/10