beads/internal/httpapi/batch_apply.go
Julian Knutsen 6cd1184687
feat(httpapi): opt-in large issues:batchApply envelope (≤1000 items, one transaction) (#7051)
* test(storage): measure ApplyBatchInTx SQL statement counts (B0)

Adds a driver.Connector-level counting/latency-injecting wrapper
(internal/storage/sqlcount) plus shared plan-request fixtures
(internal/storage/batchfixtures: mol-normalized 356/712-item shapes and a
40-item classic shape) and backend-specific pinning tests + benchmarks
(internal/storage/embeddeddolt, internal/storage/dolt) that measure the
ACTUAL number of SQL statements issueops.ApplyBatchInTx issues for each
shape, on both the embedded and Dolt (TCP) backends.

Measured totals (Prepare+Exec+Query+StmtExec+StmtQuery inside one
transaction): 356 items ~7008-7009, 712 items 14014-14015, 40 items 846,
identical on both backends (expected: statement issuance is driven by Go
application logic, not backend-specific SQL planning). Pinned with a small
(+/-2) tolerance to absorb an observed, host-contention-driven +/-1 Query
drift (Exec is deterministic on every run). This is the regression baseline
slice B2 (a later, lighter fast path) must demonstrably lower.

batchfixtures.UncappedPlan bypasses issueops.MaxApplyBatchItems so this
measurement does not depend on slice B1's cap raise.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

* feat(httpapi,issueops): raise batchApply envelope to 1000 items with bounded large-apply handling

POST /v0/beads/issues:batchApply now accepts up to 1000 items (up from 100);
issueops.MaxApplyBatchItems moves accordingly and the request body cap rises
to 16 MiB. The whole plan still applies in exactly ONE transaction. A request
over 100 items runs under a separate, flat, operator-configured whole-run
budget (`bd serve --large-apply-ceiling`, 5 minutes by default) instead of
the ordinary 60s request deadline; a request at or under 100 items is
unaffected and keeps exactly the 60s deadline and 15s commit-retry ceiling it
always had.

Concurrency and shutdown are bounded, not just documented:
- Oversized requests serialize through a dedicated one-wide semaphore, and at
  most one caller may wait for that semaphore at a time (a CAS-guarded single
  waiter) — a burst of oversized requests is shed with 503 rather than
  queuing and starving ordinary traffic behind the general semaphore.
- The run deadline is stamped only once the slot is actually acquired, not
  at the time the request arrived.
- Graceful shutdown closes new admissions before computing its drain budget,
  then waits out an in-flight large apply up to its own remaining deadline
  (plus a few seconds' grace for the response to write); any large apply
  still queued for the semaphore when shutdown begins is refused rather than
  extending the drain further. A queued caller that disconnects gets a clean
  503, never a misclassified context error. `bd serve`'s orchestrator stop
  grace must be at least `--large-apply-ceiling` — see
  engdocs/SERVE_RUNBOOK.md.
- ApplyBatch's commit-retry budget only extends past the ordinary 15s ceiling
  on this explicit large-apply path (via a context marker), with headroom
  reserved so the final retry attempt can't be killed mid-commit by the
  context's own expiry.

A new `issues.batchApplyLarge` capability token on `GET /v0/beads/context`
lets a client learn whether a given server accepts the raised envelope.

Wall-clock measurement (embedded Dolt, whole-transaction build+commit) backs
the 5-minute ceiling default: 356 items ~12.6s (23.8x headroom), 1000 items
~57.5s (5.2x headroom).

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

* fix(httpapi,serve): wire the large-apply retry budget; repair stale pins

Review fixes for the large batchApply envelope, on top of the test
repairs the PR needed against main.

The extended commit-retry budget was never switched on.
internal/storage/uow's retryTxBudget lets a large apply keep retrying a
commit-time conflict past the ordinary 15s ceiling only when its context
carries issueops.WithExtendedRetryBudget's marker, and the docs on both
sides say admitLargeApply sets it, but nothing did: the marker's only
callers were the uow tests, which mark their contexts by hand. A
1000-item apply that hit a conflict still gave up after 15s with most of
its 5-minute ceiling unused. admitLargeApply now marks the run context
it builds, and TestApplyBatchMarksOnlyALargeApplyForTheExtendedRetryBudget
pins the seam over HTTP: marked at 101 and 1000 items, unmarked at 100.
Removing the mark fails the 101 and 1000 cases; marking every request in
handleApplyBatch fails the 100 case.

Smaller fixes from the same review:

- `bd serve --large-apply-ceiling` repeated the 5m default as a literal
  of its own, and since bd serve always passes a positive ceiling, that
  literal, not httpapi's, was the default every deployment ran. Export
  httpapi.DefaultLargeApplyCeiling, default the flag to it, and pin the
  flag's 5m0s default in TestServeFlags.
- admitLargeApply's release cancels the run context before clearing
  s.largeApplyDeadline, the order acquireLargeApply's doc states; it
  cleared the deadline first.
- Drop the batchfixtures test's dead t.Skipf, which waited for the cap
  to reach 1000, and correct a graph-counter comment that still cited
  the 100-item cap and the TestServeFlags doc line that called the
  no-argument server unchanged.

The PR raises issueops.MaxApplyBatchItems from 100 to 1000 and adds
`bd serve --large-apply-ceiling`, but two base-owned tests still pinned
the old surface and went red on every lane that runs them (Bazel / test,
Test (macos-latest)):

- TestServeFlags pins the exact `bd serve` flag list; add
  large-apply-ceiling.
- TestPlanApplyBatchRefusesAnUnusableRequest/too_many_items built 101
  items and expected "at most 100 items", a request the 1000 cap now
  plans successfully. Size the over-cap request and its message from
  issueops.MaxApplyBatchItems, as TestPlanApplyBatchAcceptsTheBoundary
  already does for the other side of the cap; the cap's value stays
  pinned by TestCapBatchApplyLargeTiesAllThreeLimits.

The PR's own TestLargeApplyBurstNeverStarvesSmallRequestsForSemTimeout
also went red in PR Core (-race, -p 4 on a 4-vCPU runner): all eight
small requests took 160-165ms against a 100ms bound. The stall it
guards against lasts semTimeout (5s), while a small request can still
queue briefly behind the refused large requests, each holding a general
sem slot while its body decodes. Bound it at a second, the "want well
under a second" its sibling TestLargeApplyQueueDoesNotStarveOrdinaryRequests
already uses. Pinned to four cores beside 24 busy loops, small requests
measured 183ms mean and 338ms max, and the 1s bound passed 80/80 runs;
with the one-waiter gate in acquireLargeApply disabled the test still
fails 3/3, on its 2s never-completed check.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(uow): stop a doomed large-apply retry before ctx cuts it off

retryTxBudget sizes maxElapsed once at ApplyBatch entry, but says nothing
about how long an individual attempt takes. A slow first attempt (e.g.
157s of a large apply's commit) left little of the budget for a retry,
which then got killed mid-transaction by ctx's own deadline instead of
finishing and reporting its own outcome -- surfacing as
context.DeadlineExceeded (isUnavailable excludes it -> a generic 500)
instead of the exhausted-write-conflict result this loop already reports
correctly (-> 503 + Retry-After).

RunTxResultWithin now tracks the previous attempt's duration and, before
starting a new one, stops early with the last real serialization error
when time left on ctx is less than that duration plus a fixed headroom --
but only when publicops.HasExtendedRetryBudget(ctx) is set (the same
marker retryTxBudget itself checks). Applying the look-ahead
unconditionally regressed TestApplyBatchHonorsAnExtendedRetryBudgetFromContext's
unmarked case: ordinary requests have cheap attempts and are meant to run
until their own ctx deadline ends the loop, and the large-apply-sized 5s
headroom was tripping that for any short-ctx caller with a near-instant
failing attempt.

Adds TestRunTxResultWithin_SlowAttemptStopsBeforeADoomedRetry (a 1.2s
commit against a 2s marked ctx must stop after one attempt with its own
serialization error, never DeadlineExceeded, and never touch a second
unit of work) and its control, TestRunTxResultWithin_ShortAttemptStillRetries
(a 50ms attempt against the same shape must still retry and succeed).
Both mutation-checked against the look-ahead condition and its gate.

Also updates the stale "items accepts at most 100 entries" prose on the
applyBatch endpoint description (openapi.v0.yaml) to the large-envelope
wording already used elsewhere in the spec: 1000-item cap behind the
issues.batchApplyLarge capability, large-apply-ceiling deadline extension,
<=100 unaffected. dependencies:count's own "at most 100" lines are
unrelated and untouched.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

* fix(ci,serve): fit the 1000-item cap's tests in the storage shard budget

Review fixes for the relaunched #7051, iteration 1.

Raising issueops.MaxApplyBatchItems from 100 to 1000 made the base-owned
TestBatchApplyContract take 604s under -race, against 97s at the old cap
(main run 35862663577). It hashed onto the legacy test-embedded-storage
shard 2/5, beside ~5m of other tests, and that shard hit its 15m budget
twice (PR Risk run 36807191489). PR Risk no longer runs the legacy tier
on same-repo PRs (#7071), but Main's push run and fork PRs still do, so
the merge would have carried the timeout to Main.

Raising the jobs' timeout-minutes alone would not fix it: the shard
script passes the binary its own -test.timeout=15m. So:

- embedded-storage-test-shard.sh passes -test.timeout=20m (and its go
  test fallback -timeout 20m), as embedded-test-shard.sh does for
  cmd/bd.
- test-embedded-storage in main.yml and pr-risk.yml: timeout-minutes 15
  -> 25, the script's 20m plus setup.
- embeddeddolt_embedded_test appends -test.timeout=19m, as
  bd_embedded_test does, so Go's timeout, which dumps every goroutine,
  fires before rbe-west's 1200s action kill, which does not.
  TestBazelEmbeddedJobMirrorsEmbeddedTier now checks the override on
  both sharded targets instead of only cmd/bd's, and the PR Risk
  coverage test pins the storage target's new args.
- embedded-storage-test-shards.txt places TestBatchApplyContract on
  shard 4, which has ~3m of other tests, and the PR's two large-batch
  tests, TestLargeBatchApplyStatementCounts_Embedded (247s) and
  TestLargeBatchApplyWallClock_Embedded (261s), on shards 2 and 5. From
  the PR run's per-test times, the heaviest shard projects to ~13m, down
  from shard 2's 15m-plus.

Also correct the stop-grace guidance. For an in-flight large apply,
Serve's drain waits out the apply's remaining deadline plus drainGrace
(5s), so the orchestrator's stop grace must be at least
--large-apply-ceiling plus 5s, as Serve's own doc says. The runbook, the
flag help, the LargeApplyCeiling field doc and the CHANGELOG entry asked
for the ceiling alone.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Gas City Adopt-PR <adopt-pr@gascity.local>
Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
2026-10-01 08:32:56 -07:00

1359 lines
54 KiB
Go

package httpapi
import (
"encoding/json"
"errors"
"fmt"
"net/http"
"strings"
"time"
"unicode/utf8"
"github.com/steveyegge/beads/internal/httpapi/apigen"
"github.com/steveyegge/beads/internal/storage"
"github.com/steveyegge/beads/internal/types"
"github.com/steveyegge/beads/issueops"
)
// The wire adapter over issueops.BatchApplier: an ORDERED, heterogeneous plan
// applied as one transaction or not at all.
//
// It carries the claim's posture verbatim. The actor is caller-ASSERTED
// provenance for the audit trail and not authenticated identity; hooks do not
// fire and the per-command auto-commit machinery does not run. The only durable
// effect is the single storage commit the role makes inside its own
// transaction.
//
// EVERYTHING ABOVE THE ROLE IS ARGUMENT VALIDATION. What the values MEAN — how
// a key resolves and which way it may reach, whether a precondition holds
// as-modified, whether a waits-for gate is one of the two, what the graph the
// whole request built permits — belongs to issueops.BatchApplier. This file
// decodes a body four levels deep, refuses the shapes the document refuses, and
// maps the role's TYPED refusals onto the frozen codes.
const (
// maxApplyBatchItems is the document's cap on `items`, and the role's own
// bound restated at the edge so an over-long request costs no database work.
// It bounds how long one request may hold a write transaction — not batch
// semantics, which have no size in them.
maxApplyBatchItems = issueops.MaxApplyBatchItems
// maxApplyBatchBodyBytes bounds the request body. A thousand items each
// carrying a description, a design, acceptance criteria and a metadata
// document is the shape this has to admit, so it refuses the absurd before
// any of it is parsed.
//
// Raised from 4 MiB to 16 MiB alongside maxApplyBatchItems's 100->1000
// raise: the measured max shape's body (102 creates, descriptions
// totaling about 878 KB) is about 1.1 MB, and 1000 items at the same
// density is comfortably inside 16 MiB with headroom for a caller that
// writes longer descriptions than that sample.
//
// See TestCapBatchApplyLargeTiesAllThreeLimits: this constant,
// largeApplyItemThreshold and issueops.MaxApplyBatchItems are pinned
// together with CapBatchApplyLarge, the token that advertises all three.
//
// See maxInflight's doc comment (server.go) for the worst-case concurrent
// decode memory this implies: maxInflight requests may each be decoding a
// body up to this size at once, before the large-apply semaphore narrows
// how many of them can go on to actually run.
maxApplyBatchBodyBytes = 16 << 20
// largeApplyItemThreshold is where a request crosses from "the original,
// always-supported shape" to "the raised envelope": a request with MORE
// than this many items holds the largeApplySem slot (serializing
// oversized transactions one at a time, Server.acquireLargeApply) and
// runs under a whole separate, EXTENDED budget (Server.largeApplyCeiling)
// instead of the ordinary requestDeadline every other request gets. It is
// deliberately the OLD cap, not some fraction of the new one: every
// request at or under this threshold runs under EXACTLY the deadline it
// always has — route() applies requestDeadline unconditionally, and
// nothing in this file touches it for such a request.
largeApplyItemThreshold = 100
)
// The document's member list at each of this body's levels. Every schema is
// additionalProperties: false, so anything else is refused BY NAME — which is
// why each level is decoded as raw members first, and why the levels below the
// request are checked in the same pass that projects them.
//
// PRESENCE IS THE SIGNAL at three of these levels and that is the second reason
// for the raw decode: an item's payload members carry the tagged union's
// disagreement cases, a patch member present is written where an absent one is
// untouched, and a metadata member present holding `null` is a value.
var (
applyBatchRequestMembers = []string{"actor", "force_id_prefix", "items", "provenance", "skip_per_edge_cycle_check"}
applyItemMembers = []string{"close", "create", "dep_add", "kind", "update"}
applyRefMembers = []string{"id", "key"}
applyCreateItemMembers = []string{
"acceptance_criteria", "assignee", "defer_until", "description", "design",
"due_at", "ephemeral", "estimated_minutes", "external_ref", "id",
"issue_type", "key", "labels", "metadata", "metadata_refs", "no_history",
"notes", "owner", "priority", "sender", "status", "title",
}
applyUpdateItemMembers = []string{
"expected_assignee", "expected_status", "expected_version",
"force_assignee_transfer", "force_close_policy", "force_notes_overwrite", "patch", "target",
}
applyPatchMembers = []string{
"acceptance_criteria", "append_notes", "assignee", "defer_until",
"description", "design", "due_at", "estimated_minutes", "external_ref",
"issue_type", "labels", "metadata", "notes", "owner", "priority",
"status", "title",
}
applyLabelPatchMembers = []string{"add", "remove", "replace"}
applyMetadataPatchMembers = []string{"merge", "replace", "set", "unset"}
applyCloseItemMembers = []string{"expected_version", "force", "reason", "session", "target"}
applyDepAddItemMembers = []string{"metadata", "source", "target", "type"}
// applyItemKinds is the tag vocabulary, paired with the payload member each
// value names. It is one map rather than a switch so the enum, the member
// names and the agreement check cannot drift into three opinions.
applyItemKinds = map[issueops.ItemKind]string{
issueops.ItemCreate: "create",
issueops.ItemUpdate: "update",
issueops.ItemClose: "close",
issueops.ItemDepAdd: "dep_add",
}
// applyNullablePatchMembers is the closed set on which explicit `null`
// CLEARS rather than refuses, exactly as it is for PATCH
// /v0/beads/issues/{id}: they are the members the role models as Field[*T],
// because a pointer is the only thing a clear has to write.
applyNullablePatchMembers = map[string]bool{
"estimated_minutes": true,
"external_ref": true,
"due_at": true,
"defer_until": true,
}
)
// handleApplyBatch applies every item in the request body in order, or applies
// none of them.
//
// The transaction boundary, the ref resolution, the as-modified preconditions,
// the close policy, the assignee fence, the metadata splice and the end gate
// all belong to issueops.BatchApplier.
func (s *Server) handleApplyBatch(w http.ResponseWriter, r *http.Request) {
if !s.requireNoQuery(w, r) {
return
}
if !s.requireJSONContent(w, r) {
return
}
request, ok := s.applyBatchRequest(w, r)
if !ok {
return
}
// One-wide "large write" semaphore: a request carrying more than
// largeApplyItemThreshold items serializes against every other large
// apply, so at most one oversized transaction holds a write connection
// at a time, and runs under an EXTENDED budget built fresh at the moment
// it is admitted (Server.acquireLargeApply). An ordinary (<=threshold)
// request never touches this, is never waited on by one, and keeps
// r.Context() exactly as route() built it — requestDeadline,
// unconditionally, the same as it always has.
if len(request.Items) > largeApplyItemThreshold {
runCtx, release, err := s.acquireLargeApply(r.Context())
if err != nil {
s.failApplyBatch(w, r, request, err)
return
}
defer release()
r = r.WithContext(runCtx)
}
applier, err := s.batchApplier(r)
if err != nil {
s.failApplyBatch(w, r, request, err)
return
}
result, err := applier.ApplyBatch(r.Context(), request)
if err != nil {
s.failApplyBatch(w, r, request, err)
return
}
writeJSON(w, applyBatchResponse(result))
}
// applyBatchRequest decodes and validates the body, and reports whether the
// request may proceed. Every refusal here happens BEFORE any database work,
// which is what lets these 400s reflect the caller's own input back.
func (s *Server) applyBatchRequest(w http.ResponseWriter, r *http.Request) (issueops.ApplyBatchRequest, bool) {
refuse := func(res *Result) (issueops.ApplyBatchRequest, bool) {
s.fail(w, r, *res)
return issueops.ApplyBatchRequest{}, false
}
members, res := decodeJSONObject(w, r, maxApplyBatchBodyBytes)
if res != nil {
return refuse(res)
}
if offender, unknown := unknownMember(members, applyBatchRequestMembers); unknown {
s.failUnknownMember(w, r, offender, applyBatchRequestMembers)
return issueops.ApplyBatchRequest{}, false
}
actor, ok := s.bodyActor(w, r, members)
if !ok {
return issueops.ApplyBatchRequest{}, false
}
provenance, res := applyTextMember(members, "", "provenance")
if res != nil {
return refuse(res)
}
forceIDPrefix, res := applyBoolMember(members, "", "force_id_prefix")
if res != nil {
return refuse(res)
}
skipPerEdge, res := applyBoolMember(members, "", "skip_per_edge_cycle_check")
if res != nil {
return refuse(res)
}
items, res := applyItems(members)
if res != nil {
return refuse(res)
}
return issueops.ApplyBatchRequest{
Actor: actor,
Items: items,
Provenance: provenance,
ForceIDPrefix: forceIDPrefix,
SkipPerEdgeCycleCheck: skipPerEdge,
}, true
}
// applyItems validates `items` and projects it onto the role's items, in
// request order — which this operation never changes.
func applyItems(members map[string]json.RawMessage) ([]issueops.ApplyItem, *Result) {
refuse := func(detail string) *Result {
res := InvalidArgument("items", ReasonInvalidValue, detail)
return &res
}
raw, ok := members["items"]
if !ok {
return nil, refuse("`items` is required")
}
var rawItems []map[string]json.RawMessage
if err := json.Unmarshal(raw, &rawItems); err != nil || rawItems == nil {
return nil, refuse("`items` must be an array of objects")
}
switch {
case len(rawItems) == 0:
return nil, refuse("`items` must carry at least one item; a request that writes nothing is refused rather than answered")
case len(rawItems) > maxApplyBatchItems:
return nil, refuse(fmt.Sprintf("`items` carries %d items; the limit is %d per request", len(rawItems), maxApplyBatchItems))
}
items := make([]issueops.ApplyItem, 0, len(rawItems))
for i, rawItem := range rawItems {
item, res := applyItem(i, rawItem)
if res != nil {
return nil, res
}
items = append(items, item)
}
return items, nil
}
// applyItem projects one tagged item onto the role's item.
//
// IT ENFORCES THE TAG BY HAND, and that is the cost of the document's own
// doctrine rather than an oversight here. The item is a single-shape object
// with a required `kind` and four optional payload members, because a schema
// alternation would need a composition keyword this document does not use — so
// nothing in the generated type stops a client sending two payloads, or a
// payload the kind does not name. Both are answered here, before the role sees
// a request it would have to refuse for the same reason.
func applyItem(index int, raw map[string]json.RawMessage) (issueops.ApplyItem, *Result) {
prefix := applyItemParam(index, "")
if raw == nil {
res := InvalidArgument(prefix, ReasonInvalidValue, "an item must be a JSON object")
return issueops.ApplyItem{}, &res
}
if offender, unknown := unknownMember(raw, applyItemMembers); unknown {
return issueops.ApplyItem{}, applyUnknownMember(prefix+".", offender, applyItemMembers)
}
kindText, res := applyRequiredText(raw, prefix+".", "kind")
if res != nil {
return issueops.ApplyItem{}, res
}
kind := issueops.ItemKind(kindText)
named, known := applyItemKinds[kind]
if !known {
res := InvalidArgument(prefix+".kind", ReasonInvalidValue,
"`kind` must be one of "+strings.Join(applyKindNames(), ", "))
return issueops.ApplyItem{}, &res
}
// The tag and the payloads have to agree in both directions: a kind with no
// payload is an item that does nothing, and a payload the kind does not name
// is an item whose two halves disagree.
var present []string
for _, member := range []string{"create", "update", "close", "dep_add"} {
if _, ok := raw[member]; ok {
present = append(present, member)
}
}
switch {
case len(present) == 0:
res := InvalidArgument(prefix+"."+named, ReasonInvalidValue,
"an item of kind `"+kindText+"` must carry its `"+named+"` payload")
return issueops.ApplyItem{}, &res
case len(present) > 1:
res := InvalidArgument(prefix+"."+present[1], ReasonInvalidValue,
"an item carries exactly one payload; this one carries "+strings.Join(present, " and "))
return issueops.ApplyItem{}, &res
case present[0] != named:
res := InvalidArgument(prefix+"."+present[0], ReasonInvalidValue,
"this item is kind `"+kindText+"` but carries the `"+present[0]+"` payload; the two must name the same verb")
return issueops.ApplyItem{}, &res
}
payload, res := applyObjectMember(raw, prefix+".", named)
if res != nil {
return issueops.ApplyItem{}, res
}
item := issueops.ApplyItem{Kind: kind}
payloadPrefix := prefix + "." + named + "."
switch kind {
case issueops.ItemCreate:
create, res := applyCreateItem(payloadPrefix, raw[named], payload)
if res != nil {
return issueops.ApplyItem{}, res
}
item.Create = create
case issueops.ItemUpdate:
update, res := applyUpdateItem(payloadPrefix, raw[named], payload)
if res != nil {
return issueops.ApplyItem{}, res
}
item.Update = update
case issueops.ItemClose:
closeItem, res := applyCloseItem(payloadPrefix, payload)
if res != nil {
return issueops.ApplyItem{}, res
}
item.Close = closeItem
case issueops.ItemDepAdd:
depAdd, res := applyDepAddItem(payloadPrefix, payload)
if res != nil {
return issueops.ApplyItem{}, res
}
item.DepAdd = depAdd
}
return item, nil
}
// applyCreateItem projects a create payload onto the role's create item.
//
// It decodes into the GENERATED struct after the raw member check, which is
// what makes a member's type the DOCUMENT's type: `priority: "high"` is refused
// here rather than reaching a role that would have to guess what the caller
// meant.
func applyCreateItem(prefix string, encoded json.RawMessage, raw map[string]json.RawMessage) (*issueops.CreateItem, *Result) {
refuse := func(member, detail string) *Result {
res := InvalidArgument(applyParam(prefix, member), ReasonInvalidValue, detail)
return &res
}
if offender, unknown := unknownMember(raw, applyCreateItemMembers); unknown {
return nil, applyUnknownMember(prefix, offender, applyCreateItemMembers)
}
var wire apigen.ApplyCreateItem
if err := json.Unmarshal(encoded, &wire); err != nil {
return nil, refuse("", "a `create` member carries the wrong JSON type")
}
if strings.TrimSpace(wire.Title) == "" {
return nil, refuse("title", "`title` is required and must not be blank")
}
// The role validates the type, the status and the id prefix against the
// workspace's own configuration, which this server cannot read without a
// transaction; what is checked here is only what the schema declares. A
// SLICE and not a map, so an item breaking two rules always names the same
// offender: `param` is what a client dispatches on and it must not depend on
// map order.
for _, bounded := range []struct {
member string
value *string
}{
{"title", &wire.Title}, {"id", wire.Id}, {"key", wire.Key},
{"issue_type", wire.IssueType}, {"status", wire.Status},
{"assignee", wire.Assignee}, {"owner", wire.Owner},
{"external_ref", wire.ExternalRef}, {"sender", wire.Sender},
} {
if res := applyBoundedText(prefix, bounded.member, bounded.value); res != nil {
return nil, res
}
}
if wire.Ephemeral != nil && *wire.Ephemeral && wire.NoHistory != nil && *wire.NoHistory {
return nil, refuse("no_history", "`ephemeral` and `no_history` select different retention modes; send one")
}
issue := &types.Issue{
ID: derefString(wire.Id),
Title: wire.Title,
Description: derefString(wire.Description),
Design: derefString(wire.Design),
AcceptanceCriteria: derefString(wire.AcceptanceCriteria),
Notes: derefString(wire.Notes),
Status: types.Status(derefString(wire.Status)),
IssueType: types.IssueType(derefString(wire.IssueType)),
Assignee: derefString(wire.Assignee),
Owner: derefString(wire.Owner),
EstimatedMinutes: wire.EstimatedMinutes,
ExternalRef: wire.ExternalRef,
DueAt: wire.DueAt,
DeferUntil: wire.DeferUntil,
Sender: derefString(wire.Sender),
Metadata: wire.Metadata,
}
if wire.Priority != nil {
if *wire.Priority < 0 || *wire.Priority > 4 {
return nil, refuse("priority", fmt.Sprintf("`priority` is %d; the range is 0 to 4", *wire.Priority))
}
issue.Priority = *wire.Priority
}
if wire.Labels != nil {
if res := applyBoundedLabels(prefix, "labels", *wire.Labels); res != nil {
return nil, res
}
issue.Labels = *wire.Labels
}
if wire.Ephemeral != nil {
issue.Ephemeral = *wire.Ephemeral
}
if wire.NoHistory != nil {
issue.NoHistory = *wire.NoHistory
}
item := &issueops.CreateItem{Key: derefString(wire.Key), Issue: issue}
refs, res := applyMetadataRefs(prefix, raw)
if res != nil {
return nil, res
}
item.MetadataRefs = refs
return item, nil
}
// applyMetadataRefs decodes the one member whose keys may reach FORWARD. The
// refs are read raw rather than off the generated map so an unknown member
// inside one is refused by name like every other one.
func applyMetadataRefs(prefix string, raw map[string]json.RawMessage) (map[string]issueops.Ref, *Result) {
member, ok := raw["metadata_refs"]
if !ok {
return nil, nil
}
var entries map[string]json.RawMessage
if err := json.Unmarshal(member, &entries); err != nil || entries == nil {
res := InvalidArgument(prefix+"metadata_refs", ReasonInvalidValue,
"`metadata_refs` must be an object whose values are refs")
return nil, &res
}
refs := make(map[string]issueops.Ref, len(entries))
for key, encoded := range entries {
ref, res := applyRef(prefix+"metadata_refs."+key, encoded)
if res != nil {
return nil, res
}
refs[key] = ref
}
return refs, nil
}
// applyUpdateItem projects an update payload onto the role's update item.
func applyUpdateItem(prefix string, encoded json.RawMessage, raw map[string]json.RawMessage) (*issueops.UpdateItem, *Result) {
if offender, unknown := unknownMember(raw, applyUpdateItemMembers); unknown {
return nil, applyUnknownMember(prefix, offender, applyUpdateItemMembers)
}
// The guard is read BEFORE the whole-item decode, and through the same
// reader every other operation uses. It is the one member of this item whose
// wire type is not its Go type — a decimal string standing for an int64 —
// so letting the struct decode reach it first would answer a mistyped token
// with the item-level "wrong JSON type" instead of a 400 naming the member.
expectedVersion, res := applyVersionGuardMember(raw, prefix)
if res != nil {
return nil, res
}
var wire apigen.ApplyUpdateItem
if err := json.Unmarshal(encoded, &wire); err != nil {
res := InvalidArgument(applyParam(prefix, ""), ReasonInvalidValue, "an `update` member carries the wrong JSON type")
return nil, &res
}
target, res := applyRequiredRef(prefix, raw, "target")
if res != nil {
return nil, res
}
for _, bounded := range []struct {
member string
value *string
}{{"expected_status", wire.ExpectedStatus}, {"expected_assignee", wire.ExpectedAssignee}} {
if res := applyBoundedText(prefix, bounded.member, bounded.value); res != nil {
return nil, res
}
}
patchRaw, res := applyObjectMember(raw, prefix, "patch")
if res != nil {
return nil, res
}
patch, res := applyPatch(prefix+"patch.", raw["patch"], patchRaw)
if res != nil {
return nil, res
}
item := &issueops.UpdateItem{Target: target, Patch: patch, ExpectedVersion: expectedVersion}
if wire.ExpectedStatus != nil {
status := issueops.Status(*wire.ExpectedStatus)
item.ExpectedStatus = &status
}
item.ExpectedAssignee = wire.ExpectedAssignee
if wire.ForceClosePolicy != nil {
item.ForceClosePolicy = *wire.ForceClosePolicy
}
if wire.ForceAssigneeTransfer != nil {
item.ForceAssigneeTransfer = *wire.ForceAssigneeTransfer
}
if wire.ForceNotesOverwrite != nil {
item.ForceNotesOverwrite = *wire.ForceNotesOverwrite
}
return item, nil
}
// applyPatch projects the decoded `patch` member onto the role's IssuePatch.
//
// MEMBER PRESENCE IS THE SIGNAL, which is why this level is read as raw members
// too: a member present sets the role's Field, a member absent leaves the field
// untouched, and the generated struct cannot tell those apart because it models
// both as a nil pointer. Explicit `null` is a third state read straight off the
// raw bytes — a clear on the four nullable members, and a 400 naming the member
// everywhere else.
func applyPatch(prefix string, encoded json.RawMessage, raw map[string]json.RawMessage) (issueops.IssuePatch, *Result) {
refuse := func(member, detail string) (issueops.IssuePatch, *Result) {
res := InvalidArgument(applyParam(prefix, member), ReasonInvalidValue, detail)
return issueops.IssuePatch{}, &res
}
if len(raw) == 0 {
// A write that writes nothing is a client bug, not a no-op to answer.
return refuse("", "`patch` must carry at least one field; an update that updates nothing is refused rather than answered")
}
if offender, unknown := unknownMember(raw, applyPatchMembers); unknown {
return issueops.IssuePatch{}, applyUnknownMember(prefix, offender, applyPatchMembers)
}
// Explicit null, before any typed decode: unmarshaling null into *T yields
// nil, which is indistinguishable from the member being absent, so a null on
// a non-nullable member would otherwise slide through as "untouched" — a
// write the client asked for and the server silently dropped.
for name, value := range raw {
if isJSONNull(value) && !applyNullablePatchMembers[name] {
return refuse(name, "`"+name+"` is not nullable; omit it to leave the field unchanged")
}
}
var wire apigen.ApplyPatchBody
if err := json.Unmarshal(encoded, &wire); err != nil {
return refuse("", "a `patch` member carries the wrong JSON type")
}
for _, bounded := range []struct {
member string
value *string
}{
{"issue_type", wire.IssueType}, {"status", wire.Status},
{"assignee", wire.Assignee}, {"owner", wire.Owner},
{"external_ref", wire.ExternalRef},
} {
if res := applyBoundedText(prefix, bounded.member, bounded.value); res != nil {
return issueops.IssuePatch{}, res
}
}
patch := issueops.IssuePatch{}
set := func(name string) bool { _, present := raw[name]; return present }
if set("title") {
title := *wire.Title
if strings.TrimSpace(title) == "" {
return refuse("title", "`title` must not be blank")
}
if res := applyBoundedText(prefix, "title", &title); res != nil {
return issueops.IssuePatch{}, res
}
patch.Title = issueops.Field[string]{Set: true, Value: title}
}
if set("description") {
patch.Description = issueops.Field[string]{Set: true, Value: *wire.Description}
}
if set("design") {
patch.Design = issueops.Field[string]{Set: true, Value: *wire.Design}
}
if set("acceptance_criteria") {
patch.AcceptanceCriteria = issueops.Field[string]{Set: true, Value: *wire.AcceptanceCriteria}
}
// The role refuses both together too; refusing here keeps the 400 a
// statement about the request rather than a translated storage error.
if set("notes") && set("append_notes") {
return refuse("append_notes", "`notes` replaces the notes and `append_notes` adds to them; send one")
}
if set("notes") {
patch.Notes = issueops.Field[string]{Set: true, Value: *wire.Notes}
}
if set("append_notes") {
patch.AppendNotes = issueops.Field[string]{Set: true, Value: *wire.AppendNotes}
}
if set("priority") {
priority := *wire.Priority
if priority < 0 || priority > 4 {
return refuse("priority", fmt.Sprintf("`priority` is %d; the range is 0 to 4", priority))
}
patch.Priority = issueops.Field[int]{Set: true, Value: priority}
}
if set("issue_type") {
patch.IssueType = issueops.Field[issueops.IssueType]{Set: true, Value: issueops.IssueType(*wire.IssueType)}
}
if set("status") {
patch.Status = issueops.Field[issueops.Status]{Set: true, Value: issueops.Status(*wire.Status)}
}
if set("assignee") {
patch.Assignee = issueops.Field[string]{Set: true, Value: *wire.Assignee}
}
if set("owner") {
patch.Owner = issueops.Field[string]{Set: true, Value: *wire.Owner}
}
// The four nullable members. Set is true whenever the member is present; the
// VALUE is a nil pointer for an explicit null, which is how a clear reaches
// the role.
if set("estimated_minutes") {
patch.EstimatedMinutes = issueops.Field[*int]{Set: true, Value: wire.EstimatedMinutes}
}
if set("external_ref") {
patch.ExternalRef = issueops.Field[*string]{Set: true, Value: wire.ExternalRef}
}
if set("due_at") {
patch.DueAt = issueops.Field[*time.Time]{Set: true, Value: wire.DueAt}
}
if set("defer_until") {
patch.DeferUntil = issueops.Field[*time.Time]{Set: true, Value: wire.DeferUntil}
}
if set("labels") {
labels, res := applyLabelPatch(prefix+"labels.", raw["labels"])
if res != nil {
return issueops.IssuePatch{}, res
}
patch.Labels = labels
}
if set("metadata") {
metadata, res := applyMetadataPatch(prefix+"metadata.", raw["metadata"])
if res != nil {
return issueops.IssuePatch{}, res
}
patch.Metadata = metadata
}
return patch, nil
}
// applyLabelPatch projects the label edit, which is the FULL patch here rather
// than PATCH /v0/beads/issues/{id}'s complete replacement: a plan edits a label
// set it did not compose, and replacement would mean reading it back first.
func applyLabelPatch(prefix string, encoded json.RawMessage) (issueops.LabelPatch, *Result) {
var raw map[string]json.RawMessage
if err := json.Unmarshal(encoded, &raw); err != nil || raw == nil {
res := InvalidArgument(applyParam(prefix, ""), ReasonInvalidValue, "`labels` must be a JSON object")
return issueops.LabelPatch{}, &res
}
if offender, unknown := unknownMember(raw, applyLabelPatchMembers); unknown {
return issueops.LabelPatch{}, applyUnknownMember(prefix, offender, applyLabelPatchMembers)
}
var wire apigen.ApplyLabelPatch
if err := json.Unmarshal(encoded, &wire); err != nil {
res := InvalidArgument(applyParam(prefix, ""), ReasonInvalidValue, "a `labels` member carries the wrong JSON type")
return issueops.LabelPatch{}, &res
}
patch := issueops.LabelPatch{}
for _, edit := range []struct {
member string
value *[]string
assign func([]string)
}{
{"replace", wire.Replace, func(v []string) {
// Presence is the signal, and an EMPTY array clears every label —
// which is why this is a Field rather than a plain slice.
patch.Replace = issueops.Field[[]string]{Set: true, Value: v}
}},
{"add", wire.Add, func(v []string) { patch.Add = v }},
{"remove", wire.Remove, func(v []string) { patch.Remove = v }},
} {
if _, present := raw[edit.member]; !present {
continue
}
if edit.value == nil {
res := InvalidArgument(prefix+edit.member, ReasonInvalidValue, "`"+edit.member+"` must be an array of strings")
return issueops.LabelPatch{}, &res
}
if res := applyBoundedLabels(prefix, edit.member, *edit.value); res != nil {
return issueops.LabelPatch{}, res
}
edit.assign(*edit.value)
}
return patch, nil
}
// applyMetadataPatch projects the metadata edit.
//
// `set` IS READ RAW AND NEVER THROUGH THE GENERATED MAP, and that is a
// correctness fix rather than a shortcut: the generator models a map value as
// *MetadataValue, and encoding/json answers a JSON null against a pointer by
// setting the pointer to nil before any UnmarshalJSON runs — so a caller
// writing a key to JSON null would have it collapse into an absent key, which
// on this plane is the opposite request. The raw bytes carry the literal.
func applyMetadataPatch(prefix string, encoded json.RawMessage) (issueops.MetadataPatch, *Result) {
var raw map[string]json.RawMessage
if err := json.Unmarshal(encoded, &raw); err != nil || raw == nil {
res := InvalidArgument(applyParam(prefix, ""), ReasonInvalidValue, "`metadata` must be a JSON object")
return issueops.MetadataPatch{}, &res
}
if offender, unknown := unknownMember(raw, applyMetadataPatchMembers); unknown {
return issueops.MetadataPatch{}, applyUnknownMember(prefix, offender, applyMetadataPatchMembers)
}
patch := issueops.MetadataPatch{}
if value, present := raw["replace"]; present {
if len(raw) > 1 {
res := InvalidArgument(prefix+"replace", ReasonInvalidValue,
"`replace` writes the whole document and cannot be combined with `merge`, `set` or `unset`")
return issueops.MetadataPatch{}, &res
}
patch.Replace = issueops.Field[json.RawMessage]{Set: true, Value: applyRawCopy(value)}
}
if value, present := raw["merge"]; present {
patch.Merge = issueops.Field[json.RawMessage]{Set: true, Value: applyRawCopy(value)}
}
if value, present := raw["set"]; present {
var entries map[string]json.RawMessage
if err := json.Unmarshal(value, &entries); err != nil || entries == nil {
res := InvalidArgument(prefix+"set", ReasonInvalidValue, "`set` must be an object of metadata values")
return issueops.MetadataPatch{}, &res
}
patch.Set = make(map[string]json.RawMessage, len(entries))
for key, keyValue := range entries {
patch.Set[key] = applyRawCopy(keyValue)
}
}
if value, present := raw["unset"]; present {
var keys []string
if err := json.Unmarshal(value, &keys); err != nil || keys == nil {
res := InvalidArgument(prefix+"unset", ReasonInvalidValue, "`unset` must be an array of strings")
return issueops.MetadataPatch{}, &res
}
patch.Unset = keys
}
return patch, nil
}
// applyCloseItem projects a close payload onto the role's close item.
func applyCloseItem(prefix string, raw map[string]json.RawMessage) (*issueops.CloseItem, *Result) {
if offender, unknown := unknownMember(raw, applyCloseItemMembers); unknown {
return nil, applyUnknownMember(prefix, offender, applyCloseItemMembers)
}
target, res := applyRequiredRef(prefix, raw, "target")
if res != nil {
return nil, res
}
reason, res := applyTextMember(raw, prefix, "reason")
if res != nil {
return nil, res
}
session, res := applyTextMember(raw, prefix, "session")
if res != nil {
return nil, res
}
force, res := applyBoolMember(raw, prefix, "force")
if res != nil {
return nil, res
}
expectedVersion, res := applyVersionGuardMember(raw, prefix)
if res != nil {
return nil, res
}
// There is deliberately no `expected_status` here: a close is idempotent, so
// a guard spelled to refuse an already-closed row asks for a refusal where
// this verb answers with a no-op. See the schema.
return &issueops.CloseItem{
Target: target,
Reason: reason,
Session: session,
Force: force,
ExpectedVersion: expectedVersion,
}, nil
}
// applyDepAddItem projects a dep_add payload onto the role's edge item.
//
// The gate normalization a waits-for edge gets, and the refusal a bad gate
// earns, belong to the role: this checks only that the type IS a value the
// column can hold, never that it is a member of a known-types list, because the
// edge vocabulary is OPEN and a workspace's own type must pass.
func applyDepAddItem(prefix string, raw map[string]json.RawMessage) (*issueops.DepAddItem, *Result) {
if offender, unknown := unknownMember(raw, applyDepAddItemMembers); unknown {
return nil, applyUnknownMember(prefix, offender, applyDepAddItemMembers)
}
source, res := applyRequiredRef(prefix, raw, "source")
if res != nil {
return nil, res
}
target, res := applyRequiredRef(prefix, raw, "target")
if res != nil {
return nil, res
}
edgeType, res := applyRequiredText(raw, prefix, "type")
if res != nil {
return nil, res
}
if !types.DependencyType(edgeType).IsValid() {
res := InvalidArgument(prefix+"type", ReasonInvalidValue,
fmt.Sprintf("`type` must be 1 to %d characters", types.MaxDependencyTypeLen))
return nil, &res
}
item := &issueops.DepAddItem{
Source: source,
Target: target,
Type: types.DependencyType(edgeType),
}
// The blob travels as the bytes the caller sent. The role is the single
// definition of what it will accept, and a re-encode here would be a second
// one — the metadata compare-and-set's rule, applied to an edge.
if metadata, present := raw["metadata"]; present {
item.Metadata = string(metadata)
}
return item, nil
}
// applyRequiredRef reads one ref member that must be present.
func applyRequiredRef(prefix string, raw map[string]json.RawMessage, member string) (issueops.Ref, *Result) {
encoded, ok := raw[member]
if !ok {
res := InvalidArgument(prefix+member, ReasonInvalidValue, "`"+member+"` is required")
return issueops.Ref{}, &res
}
return applyRef(prefix+member, encoded)
}
// applyRef decodes one ref and applies the exactly-one rule the schema cannot
// state: `oneOf` is unavailable in this document, so a client can construct a
// ref with both members or neither and only the server can say no.
//
// Whether a key RESOLVES — and whether it reaches backward far enough — is the
// role's question, because only the role can see the whole request's key index.
func applyRef(param string, encoded json.RawMessage) (issueops.Ref, *Result) {
refuse := func(detail string) (issueops.Ref, *Result) {
res := InvalidArgument(param, ReasonInvalidValue, detail)
return issueops.Ref{}, &res
}
var raw map[string]json.RawMessage
if err := json.Unmarshal(encoded, &raw); err != nil || raw == nil {
return refuse("a ref must be a JSON object naming a `key` or an `id`")
}
if offender, unknown := unknownMember(raw, applyRefMembers); unknown {
return issueops.Ref{}, applyUnknownMember(param+".", offender, applyRefMembers)
}
var wire apigen.Ref
if err := json.Unmarshal(encoded, &wire); err != nil {
return refuse("a ref's `key` and `id` must be strings")
}
for _, bounded := range []struct {
member string
value *string
}{{"key", wire.Key}, {"id", wire.Id}} {
if res := applyBoundedText(param+".", bounded.member, bounded.value); res != nil {
return issueops.Ref{}, res
}
}
key, id := derefString(wire.Key), derefString(wire.Id)
switch {
case key == "" && id == "":
return refuse("a ref must name a `key` or an `id`")
case key != "" && id != "":
return refuse("a ref names a `key` or an `id`, not both")
}
return issueops.Ref{Key: key, ID: id}, nil
}
// applyBatchResponse projects the role's result onto the LEAN wire result.
//
// ItemResult.Issue is deliberately dropped. The Go contract carries a post-item
// snapshot because the completion hooks over this role hand a script the row it
// is being told about — and hooks never fire on this surface, so the snapshot
// has no consumer here and a hundred hydrated issues would be a response an
// order of magnitude larger than the request that produced it. See
// issueops.ItemResult.Issue.
func applyBatchResponse(result issueops.ApplyBatchResult) apigen.ApplyBatchResponse {
// Allocated rather than passed through: `keys` is a required member and a
// request whose creates named nothing must answer with an empty object, not
// the `null` a nil map marshals to.
keys := make(map[string]string, len(result.Keys))
for key, id := range result.Keys {
keys[key] = id
}
items := make([]apigen.ApplyItemResult, 0, len(result.Items))
for _, item := range result.Items {
wire := apigen.ApplyItemResult{
Kind: apigen.ApplyItemResultKind(item.Kind),
IssueId: item.IssueID,
Changed: item.Changed,
Revision: types.RevisionToken(item.RowVersion),
}
if item.DependsOnID != "" {
dependsOn := item.DependsOnID
wire.DependsOnId = &dependsOn
}
items = append(items, wire)
}
return apigen.ApplyBatchResponse{Keys: keys, Items: items}
}
// failApplyBatch answers a refused plan, mapping the role's TYPED refusals onto
// the frozen codes and naming the item each one came from.
//
// EVERY BRANCH READS THE ROLE'S TYPED FIELDS, never its prose. That matters more
// here than anywhere else on the surface: the request is all or nothing, so
// there is no per-item result array a client could find the offender in, and
// the `item_*` problem members are the only place it exists. They come from
// *issueops.ItemError, raised inside the transaction that refused.
//
// Nothing here quotes a role message. 4xx details on this surface reflect the
// caller's own input back rather than server internals, and the real error goes
// to the log with the request id.
func (s *Server) failApplyBatch(w http.ResponseWriter, r *http.Request, request issueops.ApplyBatchRequest, err error) {
var (
refErr *issueops.RefError
itemErr *issueops.ItemError
typeConflict *issueops.DependencyTypeConflictError
hierarchy *issueops.DependencyHierarchyConflictError
endpoint *issueops.DependencyEndpointNotFoundError
openChildren *issueops.CloseOpenChildrenError
claimed *issueops.ClaimConflictError
)
// The item wrapper is read once and shared by every branch below: a refusal
// that names an item names it whatever the inner sentinel turns out to be.
errors.As(err, &itemErr)
at := func(res Result, member string) Result {
if itemErr == nil {
// A refusal the role raised without naming an item — the request's
// own, or a wrapper that lost the identity. `param` still names the
// member that carries them all, because the document promises one on
// every 400 but the body that failed to parse.
if member != "" && res.Problem.Param == nil {
param := "items"
res.Problem.Param = &param
}
return res
}
res = res.WithBatchItem(itemErr.Index, string(itemErr.Kind), itemErr.Key, itemErr.IssueID)
if member != "" {
param := applyItemParam(itemErr.Index, string(itemErr.Kind)+"."+member)
res.Problem.Param = &param
}
return res
}
switch {
// An unresolvable ref is a *RefError and it is the one 400 that carries a
// discriminator: `declared_later` tells an ORDERING mistake from a typo, and
// it is emitted in both polarities so an absent member cannot be misread as
// false. It is matched before ErrValidation because it unwraps to it.
case errors.As(err, &refErr):
s.refusedApplyBatch(r, err)
s.fail(w, r, InvalidArgument(applyRefParam(request, refErr), ReasonInvalidValue,
applyRefDetail(refErr)).
WithBatchItem(refErr.Index, applyKindAt(request, refErr.Index), refErr.Key, "").
WithDeclaredLater(refErr.DeclaredLater))
case errors.As(err, &typeConflict):
s.fail(w, r, at(newResult(CodeDependencyExists,
"this pair already carries an edge of a different type; remove it before re-adding").
WithDependencyTypeConflict(typeConflict.ExistingType, typeConflict.RequestedType), ""))
case errors.As(err, &hierarchy):
s.fail(w, r, at(newResult(CodeDependencyCycle,
"a blocking edge against the issue's own ancestor or descendant would never clear").
WithHierarchyConflict(hierarchy.IssueID, hierarchy.BlockerID, hierarchy.BlockerIsAncestor), ""))
case errors.Is(err, issueops.ErrDependencyCycle):
// No hierarchy members: this is the plain scheduling cycle, and their
// ABSENCE is what tells a client which of the two refusals it got. It may
// come from the per-edge probe or from the END GATE, which is the only
// place an edge that is legal alone and illegal in the graph this request
// built is caught.
s.fail(w, r, at(newResult(CodeDependencyCycle,
"the plan's edges would create a dependency cycle; nothing was written"), ""))
// The edge existence refusals are 400s rather than 404s, conforming to
// POST /v0/beads/dependencies:add: an edge describes a relation rather than
// acting on a row, and its target may legitimately be an "external:"
// reference this database holds nothing for.
case errors.As(err, &endpoint):
s.refusedApplyBatch(r, err)
member, detail := "source", "an edge's source names no issue in this workspace; nothing was written"
if errors.Is(err, issueops.ErrDependencyTargetNotFound) {
member, detail = "target", "an edge's target names no issue this workspace can see; nothing was written"
}
s.fail(w, r, at(InvalidArgument("", ReasonInvalidValue, detail), member))
case errors.Is(err, issueops.ErrSelfDependency):
s.refusedApplyBatch(r, err)
s.fail(w, r, at(InvalidArgument("", ReasonInvalidValue, "an issue cannot depend on itself"), "target"))
// An occupied explicit id is a 409: the body is well-formed and stays
// well-formed, and what refuses it is STATE the client cannot see without
// reading it — so recovery is to look at that state (adopt the row, pick
// another id, or stop) rather than to fix a malformed request. The identical
// body succeeded before the id was taken. Matched before ErrValidation
// because the create path wraps both.
case errors.Is(err, storage.ErrAlreadyExists):
s.refusedApplyBatch(r, err)
s.fail(w, r, at(newResult(CodeAlreadyExists,
"a create item's `id` already names a stored row; nothing was written"), "id"))
case errors.Is(err, issueops.ErrCloseOpenChildren):
res := at(newResult(CodeNotClosable,
"an item closes an issue with open children; close them first, or send the item's force flag"), "")
if errors.As(err, &openChildren) {
res = res.WithOpenChildren(openChildren.OpenChildren)
}
s.fail(w, r, res)
case errors.Is(err, issueops.ErrCloseBlocked):
s.fail(w, r, at(closeBlockedResult(err,
"an item closes a blocked issue", "clear the blocker, or send the item's force flag"), ""))
case errors.Is(err, storage.ErrAlreadyClaimed):
res := at(newResult(CodeAlreadyClaimed,
"an update transfers work away from a live foreign owner; send `force_assignee_transfer`, or guard with `expected_assignee`"), "assignee")
if errors.As(err, &claimed) {
if claimed.Assignee != "" {
res = res.WithAssignee(claimed.Assignee)
}
if claimed.Status != "" {
res = res.WithIssueStatus(string(claimed.Status))
}
}
s.fail(w, r, res)
case errors.Is(err, storage.ErrNotesOverwrite):
s.fail(w, r, at(newResult(CodeNotesOverwrite,
"an update's `patch.notes` would replace existing non-empty notes; send `force_notes_overwrite`, or use `patch.append_notes` to preserve history"), "notes"))
case errors.Is(err, issueops.ErrVersionMismatch),
errors.Is(err, issueops.ErrStatusMismatch),
errors.Is(err, issueops.ErrAssigneeMismatch):
s.fail(w, r, applyPreconditionResult(request, itemErr, err, at))
// A target an update or a close NAMED is a resource this request failed to
// address, which is POST /v0/beads/issues:delete's 404 rather than the edge
// refusal's 400 above.
case errors.Is(err, storage.ErrNotFound):
s.fail(w, r, at(NotFound(), ""))
case errors.Is(err, storage.ErrValidation):
s.refusedApplyBatch(r, err)
s.fail(w, r, at(InvalidArgument("items", ReasonInvalidValue,
"an item was refused by this workspace's own validation; nothing was written"), ""))
default:
s.failErr(w, r, err)
}
}
// applyPreconditionResult builds the 409 for a guard that missed, naming the
// guard member and echoing what the request asked for.
//
// The expected value comes from the REQUEST rather than from a read, and the
// observed value is absent, because this operation's refusal rolled its
// transaction back: a read afterwards would describe a row the refusal never
// saw. See PreconditionFailed.
func applyPreconditionResult(request issueops.ApplyBatchRequest, itemErr *issueops.ItemError, err error, at func(Result, string) Result) Result {
res := PreconditionFailed()
switch {
case errors.Is(err, issueops.ErrVersionMismatch):
res = at(res, "expected_version")
if expected := applyExpectedVersion(request, itemErr); expected != nil {
res = res.WithExpectedVersion(*expected)
}
case errors.Is(err, issueops.ErrStatusMismatch):
res = at(res, "expected_status")
if item := applyUpdateAt(request, itemErr); item != nil && item.ExpectedStatus != nil {
res = res.WithExpectedStatus(string(*item.ExpectedStatus))
}
default:
res = at(res, "expected_assignee")
if item := applyUpdateAt(request, itemErr); item != nil && item.ExpectedAssignee != nil {
res = res.WithExpectedAssignee(*item.ExpectedAssignee)
}
}
return res
}
// applyExpectedVersion reads the version guard off the refused item. Both an
// update and a close carry one, which is why this is not applyUpdateAt's caller.
func applyExpectedVersion(request issueops.ApplyBatchRequest, itemErr *issueops.ItemError) *int64 {
item := applyItemAt(request, itemErr)
switch {
case item == nil:
return nil
case item.Update != nil:
return item.Update.ExpectedVersion
case item.Close != nil:
return item.Close.ExpectedVersion
}
return nil
}
// applyUpdateAt reads the refused item's update payload, or nil when the
// refusal named no item or named one of another kind.
func applyUpdateAt(request issueops.ApplyBatchRequest, itemErr *issueops.ItemError) *issueops.UpdateItem {
if item := applyItemAt(request, itemErr); item != nil {
return item.Update
}
return nil
}
// applyItemAt reads the refused item out of the request the handler built.
//
// It is bounds-checked rather than trusted: the index comes from the role, and
// a server that indexed a slice on a number another package computed would turn
// a contract drift into a panic on a live request.
func applyItemAt(request issueops.ApplyBatchRequest, itemErr *issueops.ItemError) *issueops.ApplyItem {
if itemErr == nil || itemErr.Index < 0 || itemErr.Index >= len(request.Items) {
return nil
}
return &request.Items[itemErr.Index]
}
// applyKindAt names the kind of the item at index, for a refusal that carries
// an index but no kind of its own.
func applyKindAt(request issueops.ApplyBatchRequest, index int) string {
if index < 0 || index >= len(request.Items) {
return ""
}
return string(request.Items[index].Kind)
}
// applyRefParam spells `param` for an unresolvable ref.
//
// RefError.Member is DIAGNOSTIC PROSE — "target", "source", or "metadata_ref
// <key>" — rather than a vocabulary, so it is mapped onto the document's own
// member names rather than published. Anything that is not one of the two
// addressing refs is a metadata ref, and the whole member is named because the
// key inside it came from the caller's own object.
func applyRefParam(request issueops.ApplyBatchRequest, refErr *issueops.RefError) string {
kind := applyKindAt(request, refErr.Index)
if kind == "" {
return applyItemParam(refErr.Index, "")
}
member := "metadata_refs"
if refErr.Member == "target" || refErr.Member == "source" {
member = refErr.Member
}
return applyItemParam(refErr.Index, kind+"."+member)
}
// applyRefDetail says which of the two key diagnoses this is, in the server's
// own words. The machine-readable half is `declared_later`.
func applyRefDetail(refErr *issueops.RefError) string {
if refErr.DeclaredLater {
return "this ref names a key declared LATER in the request; a key reaches backward only, so move the item that declares it earlier"
}
return "this ref names a key no item in the request declares"
}
// refusedApplyBatch records the real refusal for the operator. The 4xx path does
// not log by default, and the role's message is the only place the underlying
// reason survives once the response carries the server's own words.
func (s *Server) refusedApplyBatch(r *http.Request, err error) {
s.event("request_refused", "request_id", requestInfo(r.Context()).id, "error", err.Error())
}
// applyItemParam spells the `param` member for a refusal inside `items`, so a
// client dispatching on it learns WHICH item and WHICH member.
func applyItemParam(index int, member string) string {
param := fmt.Sprintf("items[%d]", index)
if member == "" {
return param
}
return param + "." + member
}
// applyParam joins a level's dotted prefix to a member, and names the LEVEL
// itself when the member is empty — a refusal about the whole object rather
// than about one of its members. Without the trim that case would spell a
// `param` ending in a dot, which is a member name no schema declares.
func applyParam(prefix, member string) string {
if member == "" {
return strings.TrimSuffix(prefix, ".")
}
return prefix + member
}
// applyUnknownMember answers an unknown member below the request level, where
// the offender's name has to be qualified by the path that reached it.
func applyUnknownMember(prefix, offender string, allowed []string) *Result {
res := InvalidArgument(prefix+offender, ReasonUnknownParameter,
"this member carries "+strings.Join(allowed, ", ")+" and nothing else")
return &res
}
// applyKindNames lists the tag vocabulary in the document's order, for a
// refusal that has to spell it.
func applyKindNames() []string {
return []string{"create", "update", "close", "dep_add"}
}
// applyObjectMember reads a required member that must be a JSON object, as raw
// members so the level below it can be checked by name.
func applyObjectMember(raw map[string]json.RawMessage, prefix, member string) (map[string]json.RawMessage, *Result) {
encoded, ok := raw[member]
if !ok {
res := InvalidArgument(prefix+member, ReasonInvalidValue, "`"+member+"` is required")
return nil, &res
}
var out map[string]json.RawMessage
if err := json.Unmarshal(encoded, &out); err != nil || out == nil {
res := InvalidArgument(prefix+member, ReasonInvalidValue, "`"+member+"` must be a JSON object")
return nil, &res
}
return out, nil
}
// applyRequiredText reads a required string member, bounded the way every
// stored string on this surface is.
func applyRequiredText(raw map[string]json.RawMessage, prefix, member string) (string, *Result) {
if _, ok := raw[member]; !ok {
res := InvalidArgument(prefix+member, ReasonInvalidValue, "`"+member+"` is required")
return "", &res
}
value, res := applyTextMember(raw, prefix, member)
if res != nil {
return "", res
}
if value == "" {
res := InvalidArgument(prefix+member, ReasonInvalidValue, "`"+member+"` must not be empty")
return "", &res
}
return value, nil
}
// applyTextMember, applyBoolMember and applyVersionGuardMember are the pure twins of
// Server.storedTextMember and Server.booleanMember, with their rules unchanged:
// an absent member is the zero value the role reads as "not supplied", an
// explicit `null` is a 400 naming the member rather than a third state, and a
// string is bounded by what storage holds and refused for control characters —
// because these values land in columns that renderers print.
//
// They are functions rather than methods because this body nests four levels
// deep, and threading a ResponseWriter through every level so each could fail
// in place would put the response machinery in the middle of the projection.
func applyTextMember(raw map[string]json.RawMessage, prefix, member string) (string, *Result) {
encoded, ok := raw[member]
if !ok {
return "", nil
}
var value *string
if err := json.Unmarshal(encoded, &value); err != nil || value == nil {
res := InvalidArgument(prefix+member, ReasonInvalidValue, "`"+member+"` must be a string")
return "", &res
}
if res := applyBoundedText(prefix, member, value); res != nil {
return "", res
}
return *value, nil
}
func applyBoolMember(raw map[string]json.RawMessage, prefix, member string) (bool, *Result) {
encoded, ok := raw[member]
if !ok {
return false, nil
}
var value *bool
if err := json.Unmarshal(encoded, &value); err != nil || value == nil {
res := InvalidArgument(prefix+member, ReasonInvalidValue, "`"+member+"` must be a boolean")
return false, &res
}
return *value, nil
}
// applyVersionGuardMember is the family's revision-token reader, and it names
// the member itself rather than taking one.
//
// Its siblings above are generic because they read many members; this one has
// read exactly one on every operation that publishes a row-version guard, so the
// member name lives in the function instead of at five call sites that could
// disagree about how to spell it.
//
// THE TOKEN IS A STRING ON THE WIRE AND AN int64 INSIDE. It is opaque and
// equality-only, and it spans the full int64 range, which a JSON number does not
// survive in an IEEE-754-double parser; responses emit it as a decimal string
// (types.RevisionToken) and this is where it comes back. STRING ONLY, with no
// number-or-string transitional spelling: this API refuses unknown members by
// name and keeps one spelling of the guard, and a numeric token accepted here
// would be a value the client rounded before it ever reached the server.
//
// prefix stays, because a batch item's guard is reported qualified by the item
// that carried it where a single operation's is spelled bare.
func applyVersionGuardMember(raw map[string]json.RawMessage, prefix string) (*int64, *Result) {
encoded, ok := raw[expectedVersionMember]
if !ok {
return nil, nil
}
refuse := func() (*int64, *Result) {
res := InvalidArgument(prefix+expectedVersionMember, ReasonInvalidValue,
"`"+expectedVersionMember+"` must be a string: the `revision` token a response carried, echoed verbatim")
return nil, &res
}
var value *string
if err := json.Unmarshal(encoded, &value); err != nil || value == nil {
return refuse()
}
parsed, err := types.ParseRevisionToken(*value)
if err != nil {
return refuse()
}
return &parsed, nil
}
// applyBoundedText applies the bounds a stored string carries wherever it is
// spelled, so every level of this body refuses the same values. A nil pointer is
// an absent member and passes.
func applyBoundedText(prefix, member string, value *string) *Result {
if value == nil {
return nil
}
refuse := func(detail string) *Result {
res := InvalidArgument(prefix+member, ReasonInvalidValue, detail)
return &res
}
switch {
case types.CheckFieldLen(member, *value) != nil:
return refuse(fmt.Sprintf("`%s` is %d characters; storage holds at most %d",
member, utf8.RuneCountInString(*value), types.MaxFieldLen))
case strings.ContainsFunc(*value, isControlChar):
return refuse("`" + member + "` must not contain control characters")
}
return nil
}
// applyBoundedLabels applies the same bound to every entry of a label list.
func applyBoundedLabels(prefix, member string, labels []string) *Result {
for i, label := range labels {
if types.CheckFieldLen("label", label) != nil {
res := InvalidArgument(prefix+member, ReasonInvalidValue,
fmt.Sprintf("`%s[%d]` is %d characters; storage holds at most %d",
member, i, utf8.RuneCountInString(label), types.MaxFieldLen))
return &res
}
}
return nil
}
// applyRawCopy copies a raw member rather than aliasing the decoded body, so
// nothing downstream can be surprised by the request buffer's lifetime.
func applyRawCopy(raw json.RawMessage) json.RawMessage {
return json.RawMessage(append([]byte(nil), raw...))
}