beads/issueops/issueops.go
Julian Knutsen 3862b03ff5
feat(http): gate await members on create and update over bd serve (#7589)
* feat(issueops): gate members on create and update, one bounds validator

The library half of serving gate await fields. Every local leg already
stores await_type, await_id, timeout and waiters on create, and await_id
on update; this adds the two missing pieces.

- IssuePatch.Waiters: a whole-value replacement of the waiter list (nil
  or empty clears, stored as []), mapped by UpdateFields and by the
  unit-of-work updateSpec, deep-copied by CloneUpdateRequest, and
  classified as a non-coordination signal on the dolt leg. Duplicates are
  stored as given; dedup stays the CLI's.
- ValidateGateFields / ValidateGatePatch with *issueops.GateFieldError
  {Field, Detail} wrapping ErrValidation, mirroring ReclaimFieldError so a
  front door can name the member. The rules are the columns' only:
  await_type <= 32 characters, await_id <= 255 characters, the JSON
  waiter list <= one TEXT column (65535 bytes), and timeout >= 0. No
  await_type allowlist, no gate-type-only rule, no timer-needs-timeout
  rule. Run from ValidatePublicCreateRequest (every create shape on every
  backend) and ValidateUpdateRequest (await_id, waiters).
- Conformance: LifecycleCreate WritesTheGateMembers / BoundsTheGateMembers,
  LifecycleUpdate PatchesTheGateMembers / BoundsTheGateMembers and
  BatchApply CarriesTheGateMembers, each bound asserted from both sides,
  wired on the dolt, embedded and uow legs and parked on the http leg on
  the live refusal rows until the wire slices land.
- httpclient: IssuePatch.Waiters is refused before the dial on both patch
  documents under a new W-IssuePatch.Waiters row (refuse-not-drop), until
  the wire publishes it.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* feat(httpapi): publish the gate members on create and update

The wire half of serving gate await fields, over the library slice.

- createIssue and the issues:batchApply create item publish await_type,
  await_id, timeout (nanoseconds, the encoding Issue.timeout reads back)
  and waiters, behind the new behavior token issues.create.gates (one token
  for both create shapes, the issues.create.defaultPriority precedent).
- updateIssue's patch and the batchApply update item's patch publish
  await_id and waiters (whole-value replacement; [] clears), behind
  issues.update.gates. await_type and timeout are not patch members: no
  local update route changes them. Neither is nullable; null is refused
  naming the member, as on title.
- The handlers only decode and project (createGateMembers,
  patchGateMembers). The wire is no stricter than the library: every bound
  is the role's *issueops.GateFieldError (await_type <= 32, await_id <=
  255, the JSON waiter list <= one TEXT column, timeout >= 0), mapped to the
  existing 400 invalid_argument with param naming the member at its level
  (await_type, patch.waiters, items[i].create.timeout,
  items[i].update.patch.await_id). The spec mirrors exactly that:
  maxLength 32/255 and minimum 0, and no minLength/maxLength/maxItems on
  waiters. Blank and duplicate waiters and any await_type reach the role
  untouched.
- Spec: the four schemas, the capability prose, and the stale sentences
  (createIssue's "await_* publishes on no operation, read or write" and the
  patch's "spec_id and await_id publish nowhere yet"). apigen regenerated
  (from the remote drift test's diff), wireshape golden regenerated:
  additive entries and sides only, wire_revision stays 2.
- Client: the two tokens are mirrored in wire.behaviorCapabilities so the
  vocabulary stays set-equal with the server's. The client still refuses
  every gate member before the dial (the client slice flips that); the write-ledger
  pending sets name the newly published members under their live refusal
  rows, and the W-CreateRequest.Issue / W-IssuePatch.AwaitID /
  W-IssuePatch.Waiters reasons, the served-leg park reasons and the S4b
  waiver now say the server publishes them and this client does not send
  them yet.

Tests: internal/httpapi/gate_fields_test.go pins member mapping on all four
bodies, clear-by-empty on both patches, null and unpublished-member refusal,
GateFieldError -> param at every nesting level, and both tokens advertised.
Mutating the handlers (dropping the mapping, the GateFieldError arms or the
tokens) turns them red.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* feat(httpclient): send the gate members behind their tokens

The client half of serving gate await fields, over the wire slice.

- Create: await_type, await_id, timeout (nanoseconds) and waiters join the
  create vocabulary on createIssue and batchApply create items (now
  twenty-four members), sent verbatim when populated. Patch: await_id (by
  presence; "" clears) and waiters (whole-value; nil/empty travels as [])
  are encoded on both updateIssue's and the batchApply update item's patch.
- Token gate before the dial: a create naming any gate member needs
  issues.create.gates, a patch naming one needs issues.update.gates. Without
  the token the request refuses with wire.NewCapabilityError naming it
  (typed ErrUnsupported{Capability}), never a dropped member. batchApply
  checks the whole plan once, so no partial plan is sent. A request naming
  no gate member never reads either token. Store.requireCapabilityBeforeDial
  is the shared helper; requirePersistenceMove now uses it.
- Bounds read back typed: the server's 400 invalid_argument/invalid_value
  whose param names a gate member at any level (await_type, patch.waiters,
  items[i].create.timeout, items[i].update.patch.await_id) is rebuilt as the
  role's *issueops.GateFieldError, the problem kept in the chain (and a batch
  refusal still becomes its *ItemError). Without this the served bounds
  contracts could not pass.
- Ledger: W-IssuePatch.AwaitID RETIRED; the library slice's W-IssuePatch.Waiters row
  deleted (it never shipped outside this branch); W-CreateRequest.Issue and
  W-CreateItem.Issue no longer cover the gate members; new
  W-CreateRequest.Gates (refuse, Capability issues.create.gates) records the
  token-gated late refusal for both create shapes. engdocs ledger
  regenerated from Ledger() (262 rows: 201 refuse, 34 degrade, 27 retired).
- Write-ledger table: the wire slice's pending sets are flipped into carried.
- Served leg: the five gate cases (LifecycleCreate Writes/Bounds,
  LifecycleUpdate Patches/Bounds, BatchApply Carries) are un-parked and run;
  the S4b waiver and served comments now say await_id is sent.

Tests: internal/httpclient/gate_fields_test.go (encode on all four bodies,
clears, old-server refusal per member naming the token with no dial, no
partial batchApply plan, plain requests unaffected on old and new servers,
GateFieldError rebuilt from the problem). Disabling the three pre-dial
checks turns the refusal tests red.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* build(bazel): auto-sync BUILD files

Applied from the bazel-sync-patch artifact of https://github.com/gastownhall/beads/actions/runs/38841144413
(`make bazel-sync`: gazelle, tools/bazel/go_srcs.py, bazel mod tidy).
Only BUILD.bazel, MODULE.bazel and MODULE.bazel.lock are ever pushed; see
scripts/bazel-autofix-push.sh.

* docs(issueops): say the gate timeout rule is no column's

The gate validator's doc said its rules were the stored columns' and that
no command line can produce a negative timeout. The widths are the
columns' (VARCHAR(32), VARCHAR(255), TEXT), but timeout_ns is a BIGINT,
so a non-negative timeout is a rule no column enforces; the wire
publishes it as minimum 0. The raw store creates (CreateIssue,
CreateIssues) and import do not run the validator, so bd gate create
--timeout=-5m, a formula step's gate.timeout and a template clone still
store a negative timeout until they move onto the role.

Also say that GateFieldError.Detail is prose whose HTTP copy ends with
the server's own clause, so a caller dispatches on Field, and why a
create omits an empty waiter list while a patch sends [] to clear it.

Comments only; no behaviour change.

Agent-Signature: claude-code-claude-opus-5-5-max on behalf of Julian Knutsen
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* test(corpus): gh:run await_id persist refuses by the missing gate token

#7604 (now on main) made bd gate check persist a gh:run gate's discovered
run ID through Lifecycle.Update. Its served test expected the HTTP
client's W-IssuePatch.AwaitID refusal, which persistGateCheckAwaitID
rewords to name await_id. This PR retires that row: the patch carries
await_id behind issues.update.gates. Through the reference proxy, whose
server advertises no gate token, the persist is now refused before the
dial by the capability error ("bd serve at ... does not advertise
capability "issues.update.gates", which updateIssue requires"), so the
merged tree failed the subtest's await_id and updateIssue assertions.

The subtest now expects that refusal: no writes, the error names the
gate, the missing token and the server, no role or store method leaks,
and await_id is untouched. persistGateCheckAwaitID already returns a
capability error as the role raised it (#7604 pins that), so no code
changes. Against a server that advertises the token the persist goes
through: a scratch probe against the in-process server recorded
await_id=4242 and closed the gate.

Also correct the gate.go comment and the CHANGELOG entry that still said
the HTTP client refuses every await_id patch.

Agent-Signature: claude-code-claude-opus-5-5-max on behalf of Julian Knutsen
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: Eddie the Engineer <adopt-pr@gascity.com>
2026-10-16 17:02:27 +00:00

596 lines
30 KiB
Go

// Package issueops declares the public values used for guarded issue mutations.
//
// # Stability
//
// EXPERIMENTAL. This package is not yet covered by the project's compatibility
// promise and may change in a minor release. Pin an exact beads version.
package issueops
import (
"context"
"encoding/json"
"time"
"github.com/steveyegge/beads/internal/types"
)
// Issue is the canonical Beads issue model.
type Issue = types.Issue
// Status is an issue lifecycle status.
type Status = types.Status
// IssueType classifies an issue.
type IssueType = types.IssueType
// PersistenceMode selects the complete persistence state for an issue.
type PersistenceMode = types.PersistenceMode
// Status values accepted by issue operations.
const (
StatusOpen = types.StatusOpen
StatusInProgress = types.StatusInProgress
StatusBlocked = types.StatusBlocked
StatusDeferred = types.StatusDeferred
StatusClosed = types.StatusClosed
StatusPinned = types.StatusPinned
StatusHooked = types.StatusHooked
)
// Persistence mode values accepted by issue operations.
const (
PersistenceModePersistent = types.PersistenceModePersistent
PersistenceModeEphemeral = types.PersistenceModeEphemeral
PersistenceModeNoHistory = types.PersistenceModeNoHistory
)
// Waits-for gate values accepted by issue operations.
const (
WaitsForAllChildren = types.WaitsForAllChildren
WaitsForAnyChildren = types.WaitsForAnyChildren
)
// Field distinguishes an omitted value from an explicitly supplied zero value.
type Field[T any] struct {
// Set reports whether Value was supplied.
Set bool
// Value is the supplied value, including a zero or nil value.
Value T
}
// LabelPatch describes ordered label replacement and incremental edits.
// Operations apply Replace, then Add, then Remove, so removal wins when the
// same label appears in more than one edit.
//
// The VALUES an edit carries answer to the same field rules a create's labels
// do: a label longer than the column is refused with ErrFieldTooLong and the
// whole update writes nothing. Repetition is free in both directions — the
// same label named twice in one edit is applied once, and removing a label the
// issue does not carry is a no-op that reports Changed false.
//
// An EMPTY-STRING entry is DROPPED — not written, and not refused. A label
// row carrying "" renders as nothing and matches nothing, so writing one only
// stores junk; refusing the whole update would let one stray entry fail an
// otherwise-good multi-label edit. Dropping is a no-op rather than a partial
// write: an Add carrying only "" changes nothing and reports Changed false.
// An empty entry already stored is left alone by Add and Remove and cleared by
// a Replace that omits it, which is the ordinary set behavior and not a
// migration.
type LabelPatch struct {
// Add contains labels to add after any replacement.
Add []string
// Remove contains labels to remove after replacement and addition.
Remove []string
// Replace supplies the starting complete label set when Set is true.
Replace Field[[]string]
}
// MetadataPatch describes metadata replacement or ordered incremental edits.
// Replace is mutually exclusive with Merge, Set, and Unset. Without Replace,
// operations apply Merge, then Set keys in deterministic order, then Unset, so
// unsetting a key wins over setting or merging that key.
type MetadataPatch struct {
// Replace replaces the complete metadata document when Set is true. A nil or
// empty Value clears metadata; a nonempty Value must be valid JSON.
//
// CLEARING STORES THE EMPTY JSON DOCUMENT. An issue's metadata is never
// SQL NULL, and it is never NULL at creation either, so "created with no
// metadata" and "given metadata and then cleared" are the SAME stored
// value rather than two states for one user-visible fact. A consumer
// filtering on `metadata IS NULL` therefore matches neither, and one
// asking for an empty document matches both.
//
// Read back through a result or the Reader role, that same empty document
// arrives as an EMPTY json.RawMessage rather than the two bytes of an
// empty object — so a caller must treat nil, empty and {} as one value on
// the way out, exactly as this field treats them on the way in.
Replace Field[json.RawMessage]
// Merge merges a metadata document into the current value. A nil or empty
// Value is invalid; a nonempty Value must be a JSON object.
Merge Field[json.RawMessage]
// Set writes individual metadata keys in deterministic key order.
Set map[string]json.RawMessage
// Unset removes individual metadata keys after all other edits.
Unset []string
}
// IssuePatch describes fields that can change on an issue.
type IssuePatch struct {
Title Field[string]
Description Field[string]
Design Field[string]
AcceptanceCriteria Field[string]
// Notes replaces the notes and is mutually exclusive with AppendNotes.
Notes Field[string]
// AppendNotes appends to the notes and is mutually exclusive with Notes.
AppendNotes Field[string]
SpecID Field[string]
// AwaitID replaces the gate's condition identifier; the empty string
// clears it. Its bounds are ValidateGateFields' (GateFieldError).
AwaitID Field[string]
// Waiters is a WHOLE-VALUE replacement of the gate's waiter list; a set
// nil or empty Value clears it. The role stores the list as given:
// duplicates are kept, as local storage always has, and deduplication is
// the caller's (bd gate add-waiter dedups before it writes). Its bounds
// are the column's (GateFieldError naming GateFieldWaiters).
Waiters Field[[]string]
// Status sets the issue's status. A status that crosses from outside the
// configured done category into it answers to close policy: the update
// refuses with CloseOpenChildrenError or ErrCloseBlocked unless
// UpdateRequest.ForceClosePolicy is set. A done-to-done change and a move
// out of the done category are unaffected. Close and Reopen remain the
// operations that carry the full lifecycle policy — a crossing update
// applies the close policy, not the close semantics.
Status Field[Status]
Priority Field[int]
IssueType Field[IssueType]
Assignee Field[string]
Owner Field[string]
ClosedBySession Field[string]
EstimatedMinutes Field[*int]
ExternalRef Field[*string]
DueAt Field[*time.Time]
DeferUntil Field[*time.Time]
// Persistence selects the complete persistence state. It is unchanged when
// unset; a set value must be a known PersistenceMode. A same current mode is
// a representation-preserving no-op. Every aggregate move is atomic.
// Persistent preserves an existing durable unversioned class; promotion
// from a wisp selects normalized versioned storage. Ephemeral and NoHistory
// select wisp retention modes, never durable unversioned storage. Command
// adapters translate their own flag spelling before constructing this
// request.
Persistence Field[PersistenceMode]
// ParentID is unchanged when unset. A set empty value removes all outgoing
// parent-child edges. A set nonempty value atomically replaces all parents
// with exactly that target and does not inherit labels.
ParentID Field[string]
Labels LabelPatch
Metadata MetadataPatch
}
// CreateDependency describes a dependency created with an issue.
type CreateDependency struct {
TargetID string
Type types.DependencyType
// Reverse writes the dependency from TargetID to the new issue.
Reverse bool
Metadata string
// ThreadID associates this dependency with a discussion thread.
ThreadID string
}
// WaitsFor describes the typed metadata for a DepWaitsFor dependency. It
// records a readiness primitive; it does not define scheduling or execution
// policy.
type WaitsFor struct {
// SpawnerID is required and identifies the dependency target whose children
// are observed.
SpawnerID string
// Gate selects the readiness condition. Empty defaults to
// WaitsForAllChildren; otherwise only the exported waits-for gate constants
// are valid.
Gate string
}
// CreateRequest describes one atomic issue creation.
type CreateRequest struct {
Actor string
// Issue is included in the request's single deep snapshot at method entry;
// every transaction attempt clones that same entry snapshot. Create accepts
// its ID, content, workflow, assignment, timestamps, scheduling,
// external-reference, metadata, wisp, storage-class, marker, gate, source,
// molecule, work-type, and event fields using the canonical create rules. It
// ignores ContentHash, RowVersion, lease state, compaction state, routing
// overrides, hydration flags, and derived fields. Labels are authoritative.
// Issue.Comments and Issue.Dependencies must be empty; supply edges through
// the request's own Dependencies field. An empty Issue.CreatedBy defaults to
// Actor (the same rule for every create shape, batch items included).
Issue *Issue
// ParentID creates a typed DepParentChild edge. It must not duplicate an
// explicit edge in Dependencies.
ParentID string
// InheritLabelsFromParent copies the parent's labels at creation.
InheritLabelsFromParent bool
// Dependencies is the authoritative set of explicit edges created with the
// issue. Duplicate source-target pairs are rejected. A different type for an
// existing pair is rejected with DependencyTypeConflictError rather than
// silently retyping the edge.
Dependencies []CreateDependency
// WaitsFor creates a typed DepWaitsFor edge. It must not duplicate an
// explicit edge in Dependencies.
WaitsFor *WaitsFor
// ForceIDPrefix permits an explicit ID outside the configured prefix.
ForceIDPrefix bool
// IDPrefix is the prefix an explicit Issue.ID is checked against, and it
// OVERRIDES the one the substrate has configured. Empty means "use the
// substrate's", which is the ordinary case.
//
// It exists because the workspace's own config.yaml `issue-prefix` wins
// over the database's (GH#4957) and no implementation can see config.yaml
// — a shared server's database knows only its own prefix. Resolving that
// is the front door's job, the same way DeleteRequest.IDs are exact
// because resolution belongs to the front door; the role enforces what it
// is handed. Without it the two `bd create` routes disagree about which
// ids a workspace may mint.
IDPrefix string
// DefaultPriority says the caller has no priority for the issue, and the
// create stores DefaultCreatePriority. It exists because Issue.Priority
// cannot say "absent": its zero value is P0, a real and urgent priority, so
// a caller that leaves it unset asks for critical work. The default is
// applied here, once, for every route — the CLI's `bd create` without
// --priority and an HTTP create without `priority` both set this rather
// than spelling the number themselves.
//
// Issue.Priority must be zero when it is set; a request that sets both is
// ErrValidation, because the two would disagree about what was asked for.
DefaultPriority bool
}
// DefaultCreatePriority is the priority a create stores when its request sets
// DefaultPriority (CreateRequest, BatchCreateItem, CreateItem): P2, medium.
const DefaultCreatePriority = 2
// UpdateRequest describes an issue update.
type UpdateRequest struct {
Actor, IssueID string
Patch IssuePatch
// Claim atomically claims the issue for Actor. It first sets Assignee to
// Actor and Status to StatusInProgress, then applies Patch; a set
// Patch.Assignee or Patch.Status overrides the corresponding claim value.
// An issue is eligible when its status is built-in StatusOpen or a configured
// active status and it is unassigned, assigned to Actor, or assigned to a
// configured pool. StatusInProgress assigned to the same Actor is an
// idempotent success. A foreign assignment returns ErrAlreadyClaimed; an
// ineligible status returns ErrNotClaimable.
// ExpectedAssignee and ExpectedStatus must be nil when Claim is true because
// the claim has its own compare-and-set eligibility rules.
Claim bool
// ForceAssigneeTransfer bypasses only a genuine transfer away from a foreign
// live in-progress owner. Reasserting the exact current foreign assignee is
// idempotent and needs no force. Assignees configured as claim.pools aliases
// are transferable without force. The zero value enforces the fence. It has
// no effect with Claim or an omitted Patch.Assignee and must be false with
// ExpectedAssignee. It never bypasses other guards — in particular it is
// independent of ForceClosePolicy, and a request that sets it without
// Patch.Assignee is invalid. A command adapter with a single --force flag
// therefore maps that flag to both fields, conditioning this one on an
// assignee edit.
ForceAssigneeTransfer bool
// ForceNotesOverwrite bypasses only the refusal on a Patch.Notes that is a
// NotesReplacement of the existing notes — replacing non-empty notes with
// different non-empty content; an explicit clear is not fenced. The zero
// value enforces the fence. It has no effect without such a Patch.Notes edit, and
// a request that sets it with no Patch.Notes is invalid. It is independent
// of ForceAssigneeTransfer and ForceClosePolicy — neither bypasses this, nor
// does this bypass either of them — and, deliberately, of ExpectedAssignee
// too: a caller pairing a notes edit with an assignee compare-and-set still
// needs this bit to overwrite notes, and setting it never touches the
// assignee guard. A command adapter with a single --force flag therefore
// maps that flag to all three fields, conditioning this one on a
// Patch.Notes edit.
ForceNotesOverwrite bool
// ForceClosePolicy bypasses only close policy — the open-children refusal
// and the live-direct-blocker refusal — for a Patch.Status that crosses into
// the done category. The zero value enforces the policy. It has no effect
// without such a status change, and never bypasses validation,
// ExpectedVersion, ExpectedAssignee, ExpectedStatus, or the assignee fence.
// It is the update-side spelling of CloseRequest.Force; a command adapter
// that maps one flag to both spells both.
ForceClosePolicy bool
// AllowTemplate says the caller edits a template deliberately, so the
// template read-only refusal (*TemplateReadOnlyError) stands down for this
// request. bd label and bd set-state set it to keep their pre-guard
// behavior: they have always written to templates. The zero value
// enforces the guard. It bypasses nothing else — validation, the
// preconditions, the assignee and notes fences and close policy still
// apply — and has no effect on an issue that is not a template.
AllowTemplate bool
// The three Expected* guards below are this package's FOUNDING spelling of
// the compare-and-set family, and the family's rules are stated once, at
// length, on DeleteRequest.ExpectedVersion (deleter.go) and
// ApplyUpdateItem's guards (batchapplier.go). They hold here unchanged and
// are not restated:
//
// - THE TOKEN IS OPAQUE AND COMPARED FOR EQUALITY ONLY. RowVersion is
// minted afresh on every lifecycle write rather than incremented, so
// "newer" and "older" are not questions a version guard can answer.
// - IT GUARDS THE ROW'S LIFECYCLE STATE, NOT ITS GRAPH. RowVersion is
// reminted by status, assignee and started_at writes and deliberately
// not by label, dependency, rename or is_blocked writes, so a matching
// token does not promise the row's edges are the ones the caller saw.
// - nil DISABLES THE CHECK. Each is a pointer so that "do not check" is
// distinct from a caller requiring the never-written version 0, or the
// empty assignee, which is a real guard meaning "expected unassigned".
// - GUARDS PRESENT TOGETHER MUST ALL HOLD, and each is read inside the
// writing transaction, so there is no internal TOCTOU.
// - A MISS REFUSES AND WRITES NOTHING, with the typed sentinel for the
// guard that missed: ErrVersionMismatch, ErrStatusMismatch,
// ErrAssigneeMismatch. That is the opposite of MetadataCAS, whose miss
// is a result rather than an error; metadatacas.go states why.
//
// ExpectedVersion requires the current row version to match before the claim
// and patch. It is an independent precondition and may be combined with Claim.
ExpectedVersion *int64
// ExpectedAssignee requires the current assignee to match. A match authorizes
// the requested Patch.Assignee transfer: this compare-and-set replaces the
// ordinary anti-steal fence. It must be nil with Claim, and
// ForceAssigneeTransfer must be false when it is non-nil.
//
// THE COMPARISON IS SEPARATOR-INSENSITIVE, exactly as
// ReleaseRequest.ExpectedAssignee documents: a run of ".", "_" or "-"
// matches any other such run — except an exact "--" run, which decodes to
// "/" as gascity's rig-qualified agent encoding, so "a--b" matches "a/b"
// rather than "a__b" — and nothing else is forgiven.
ExpectedAssignee *string
// ExpectedStatus requires the current status to match. It must be nil when
// Claim is true.
ExpectedStatus *Status
// IssuePlaneOnly restricts the update to the issue plane: an ID that names
// a wisp is ErrNotFound rather than an ephemeral row to update. The zero
// value keeps the both-plane auto-resolve every caller gets today.
//
// Resolution happens INSIDE the transaction the update runs in, so a
// caller cannot be handed the plane an earlier read saw. It exists for the
// surface that serves durable issues only — refusing there rather than
// carrying a private copy of the resolve is what keeps the two front doors
// on one implementation.
IssuePlaneOnly bool
// Provenance labels the version-control history entry this mutation
// records, naming the surface the update came from. Empty selects the
// implementation's default label.
//
// It NEVER changes WHETHER history is recorded — only how the entry reads.
// An update that records none records none with it set, one that records
// one records one with it empty, and an implementation with no history to
// label ignores it. Persistence, not this, is what decides whether an issue
// is versioned at all.
Provenance string
}
// CloseRequest describes an issue closure.
type CloseRequest struct {
Actor string
IssueID string
// Reason records why the issue was closed. It is stored on the issue and
// read back on CloseResult.Issue.CloseReason, alongside the ClosedAt the
// same close stamps.
//
// THE FIRST CLOSE WINS. A second Close of an already-closed issue is the
// no-op CloseResult.Changed describes, so it writes neither this nor
// Session and the stored pair keeps the values the first close gave them
// — a replayed close cannot rewrite the record of why the work ended. A
// Reopen clears both, because they describe a closure that no longer
// holds.
Reason string
// Session names the working session that closed the issue. It is stored
// and read back on CloseResult.Issue.ClosedBySession, under the same
// first-close-wins rule as Reason.
Session string
// Force bypasses blocker and open-child close policy and the two close
// guards a force has always waived in bd: the pin, and the assignee
// authority fence. It never bypasses the template guard, validation,
// ExpectedVersion, or lifecycle rules.
Force bool
// ExpectedVersion requires the current row version to match, and is checked
// BEFORE the idempotent close — so a re-close of an already-closed issue
// still refuses on a stale token rather than reporting the no-op it would
// otherwise be. That ordering is the whole difference from composing a read
// with a close, and a conformance case pins it.
//
// The rest of the family's rules are UpdateRequest.ExpectedVersion's,
// unchanged: the token is opaque and compared for equality only, it guards
// the row's lifecycle state and not its graph, nil disables the check, and
// a miss refuses with ErrVersionMismatch having written nothing. Force
// bypasses close policy and never this guard.
ExpectedVersion *int64
// AutoCloseMolecule closes the issue's molecule root IN THE SAME
// TRANSACTION when this close completed it: the root closes itself when
// complete (AutoClosesWhenComplete), it is open, and every step under it
// is now closed (CompletedMolecule). The root close is unforced, guarded
// on the root revision read inside that transaction, and records
// MoleculeAutoCloseReason and this request's Session.
//
// It runs for an idempotent re-close too: the rule is state-derived, so a
// replayed close of the last step heals a root a crash left open.
//
// IT IS OPT-IN because an orchestrator may run its own molecule
// lifecycle (gc closes molecule roots itself); the zero value never
// touches a root. A root close its close policy refuses (a live blocker
// on the root, open children) or a close guard refuses (a template root, a
// pinned root, a root assigned to someone other than Actor — the root
// close is unforced, so each guard holds as it does for a direct close)
// leaves the root open and is reported on
// CloseResult.MoleculeAutoCloseRefusal (IsMoleculeAutoCloseRefusal); it
// never fails the step close.
//
// ANY OTHER FAILURE OF THE ROOT CLOSE FAILS THE STEP'S CLOSE. A storage
// error is returned and rolls the step's close back with it, so a retry
// re-runs both. (The CLI used to close the root after the step's close had
// committed and only warned when that failed, leaving a complete
// molecule's root open.) A write to the root that races this close is not
// such a failure: the two conflict at commit, and the backend's retry
// decides the root again (MoleculeRootClose).
//
// Two closes that finish a molecule's last steps concurrently do not
// strand its root: each close that decides a root first rewrites one
// coordination cell for it, so their transactions conflict at commit and
// the one that retries reads the other's close and closes the root.
//
// THE CELL SERIALIZES SIBLING CLOSES. Every close under an open
// auto-closing root rewrites it, whether or not that close completes the
// molecule, so any two concurrent closes of steps under the same root
// conflict and one of them retries. Under heavy fan-out a close can spend
// its whole retry budget conflicting and then fails with the
// serialization error. Closes under different roots contend only when
// their roots hash to the same cell.
AutoCloseMolecule bool
}
// ReopenRequest describes an issue reopening.
type ReopenRequest struct {
Actor string
IssueID string
// Reason records why literal closed and configured done statuses move to
// open. Non-done statuses are unchanged.
//
// WHERE it is recorded is the issue's EVENT HISTORY, carried on the
// reopened entry the move itself records. It is not stored on a field of
// the issue and ReopenResult does not surface it, so a caller that wants
// it back reads the issue's events — the same place the Actor attribution
// for the reopen lives. A reopen with no reason still records that entry;
// it simply carries none.
Reason string
// ExpectedVersion requires the current row version to match and is checked
// before a non-done no-op.
ExpectedVersion *int64
// Provenance labels the version-control history entry this reopen records,
// naming the surface it came from. Empty selects the implementation's
// default label, and the implementations do not agree on what that is — the
// store-backed ones write "bd: reopen issue" and the unit-of-work one
// "reopen issue", neither naming the issue. A caller that wants one entry
// no matter which backend served it spells one, exactly as
// UpdateRequest.Provenance describes.
//
// It NEVER changes WHETHER history is recorded — only how the entry reads.
Provenance string
}
// CreateResult reports the created issue as a detached snapshot with labels,
// dependencies, and persisted comments.
type CreateResult struct {
// Issue is a detached snapshot with labels, dependencies, and persisted
// comments.
Issue *Issue
}
// UpdateResult reports the post-update issue as a detached post-state snapshot
// with labels and dependency records. Comments are omitted.
type UpdateResult struct {
// Issue is a detached post-state snapshot with labels and dependency records.
// Comments are omitted.
Issue *Issue
// Changed reports whether the request persisted a semantic mutation. It is
// false for same-value patches and no-op updates.
Changed bool
}
// CloseResult reports the post-close issue as a detached post-state snapshot
// with labels and dependency records. Comments are omitted.
type CloseResult struct {
// Issue is a detached post-state snapshot with labels and dependency records.
// Comments are omitted.
Issue *Issue
// Changed reports whether closing persisted a semantic mutation. It is false
// for an idempotent re-close.
Changed bool
// OpenChildren is the number of open children observed by a forced close.
// It is reported even for an idempotent re-close.
OpenChildren int
// AutoClosedMolecule is the molecule root CloseRequest.AutoCloseMolecule
// closed in this close's transaction, as a post-close snapshot with
// labels. Nil when nothing was auto-closed.
AutoClosedMolecule *Issue
// MoleculeAutoCloseRefusal says why a completed molecule's root stayed
// open: its close policy or a close guard refused it
// (IsMoleculeAutoCloseRefusal), in that refusal's own message, the same on
// every backend. Empty otherwise.
MoleculeAutoCloseRefusal string
}
// ReopenResult reports the post-reopen issue as a detached post-state snapshot
// with labels and dependency records. Comments are omitted.
type ReopenResult struct {
// Issue is a detached post-state snapshot with labels and dependency records.
// Comments are omitted.
Issue *Issue
// Changed reports whether reopening persisted a semantic mutation. It is
// false when non-done statuses are unchanged.
Changed bool
}
// Lifecycle describes guarded issue mutations. A new capability gets a new
// role interface and its own accessor; never append a method here.
//
// EVERY VERB HERE MAINTAINS BLOCKED STATE under BlockedStateInvariant: a
// Create that carries edges, and an Update or Close or Reopen that moves a
// status across the closed/pinned boundary, leave the column settled for every
// row their mutation could have affected before the transaction commits.
//
// Deterministic request validation failures match ErrValidation; when a
// more-specific validation sentinel applies, it remains matchable too.
// Implementations never mutate
// caller-owned request values. Callers must not concurrently mutate
// request-owned mutable values until the call returns; snapshotting does not
// make Go data races safe. Implementations snapshot all request inputs once at
// method entry, deep-cloning caller-owned mutable values into that snapshot.
// They deep-clone that snapshot before every transaction attempt, including
// retries, and apply validation, defaulting, and normalization only to
// attempt-local clones. Result values are unspecified when error is non-nil.
// Refusals and deterministic validation failures leave persistent state
// unchanged. Other operational or commit-finalization errors can have an
// indeterminate durable outcome; callers must reread state before retrying.
type Lifecycle interface {
// Create validates and commits the complete request as one atomic mutation.
// It is create-only across issues and wisps: an occupied ID returns
// ErrAlreadyExists and leaves persistent state unchanged. A refusal or
// validation error also leaves no partial persistent state.
Create(context.Context, CreateRequest) (CreateResult, error)
// Update validates guards and commits the complete request as one atomic
// mutation. A Patch.Status that crosses from outside the configured done
// category into it answers to close policy: an unforced crossing with open
// children returns CloseOpenChildrenError and one with a live direct blocker
// returns ErrCloseBlocked, both without mutation. ForceClosePolicy bypasses
// those two refusals and nothing else. A Claim that loses its
// compare-and-set returns *ClaimConflictError carrying the state that beat
// it. An update of a template returns *TemplateReadOnlyError whatever its
// patch and force flags, unless it sets AllowTemplate — templates are
// read-only — checked after the
// compare-and-set preconditions, so a stale guard still reports the
// mismatch. A refusal or validation error leaves persistent state
// unchanged.
Update(context.Context, UpdateRequest) (UpdateResult, error)
// Close validates guards and commits the complete request as one atomic
// mutation. It moves the issue to literal StatusClosed, including from a
// configured done status. ExpectedVersion is checked first, including for an
// idempotent close. A close of an issue that is not already closed answers
// to the close guards before close policy: a template returns
// *TemplateReadOnlyError, forced or not; unforced, a pinned issue returns
// *PinnedError and one assigned to someone other than Actor (compared
// separator-insensitively) returns *CloseNotAssigneeError, matching
// ErrNotOwner. An unforced close with open children returns
// CloseOpenChildrenError without mutation. Force bypasses the pin, the
// assignee fence, and blocker and open-child policy, and reports
// OpenChildren, including for an idempotent re-close. A refusal or
// validation error leaves persistent state unchanged.
Close(context.Context, CloseRequest) (CloseResult, error)
// Reopen validates guards and commits the complete request as one atomic
// mutation. It moves literal StatusClosed and configured done statuses to
// StatusOpen; non-done statuses unchanged. A refusal or validation error
// leaves persistent state unchanged.
Reopen(context.Context, ReopenRequest) (ReopenResult, error)
}
// ErrFieldTooLong preserves the canonical model sentinel identity for a value
// that exceeds its field's maximum length. The rest of the operation error
// vocabulary is declared in errors.go.
var ErrFieldTooLong = types.ErrFieldTooLong