Files
archie/internal/domain/workflow/diff_rules.go
T
sam 8d05f99526 feat(workflow): wire pr-review into archie's own PRs before opening
StagePRReviewAndOpenPR runs the pr-review pipeline's decision phases
(1-8) against the task's own uncommitted change, using a new
localPRSource that reads Diff/Snapshot straight from Trees instead of
fetching an external PR -- there is no PR number yet to fetch one by,
which is why this trigger needed its own PRSource rather than the
forge-backed one afbk.7 already built for external PRs.

An unchallenged blocking finding parks the task with the findings
instead of opening a PR; a pipeline stage failure parks it the same
way any stage failure does. Once the PR opens, advisory findings post
against it -- best-effort, matching the deleted StagePostReviewComments'
convention: a posting failure is logged, not reverted into the outcome
the PR's existence already earned.

Implement() now uses this in place of the plain StageOpenPR. Bootstrap
and TDD are untouched -- the PRD names "the implement workflow"
specifically; extending to the others is a separate call.

Resolves this bead's first blocker (confirmed via read: CompilePinnedWorkflow
returns the same workflow.Workflow type and runs through the same
workflow.Run either compilation path, so a stage setting Outcome ends
the run identically regardless -- stagePROutput's posting logic had to
move inside StageOpenPR's own Run body, not follow it as a separate
Stage, which OpenPR's own doc comment already anticipated).

Still open on this bead: watched-repositories and operator triggers,
both blocked on the credential path for PRSource's real forge-backed
implementation reaching the sandboxed archie-agent process -- see the
bead comment for the concrete design (egress credential-capability
grant, narrow forge diff/archive HTTP methods) not yet built.
2026-09-28 13:06:42 +10:00

397 lines
15 KiB
Go

package workflow
import (
"bytes"
"context"
"errors"
"fmt"
"regexp"
"strings"
"gopkg.in/yaml.v3"
"github.com/samcharles93/archie-core/internal/domain/workflow/prreview"
)
// DiffRulesStepName is the registered workflow step type that applies a
// repository's declarative rules to the committed diff.
//
// It is the supported migration target for the deleted .archie/gate.go hook.
// That hook was repository-authored Go, discovered in the worktree and
// interpreted in-process, exporting
//
// func Check(gate.GateContext) []gate.Finding
//
// where each Finding carried a level ("error" blocked, "warn" was advisory) and
// an optional file and line. Its useful behaviour -- project-specific rules
// that shell commands cannot express, reported against the lines a change
// introduces -- is what this step type carries, with the rules as data an
// operator writes in a stored workflow definition rather than code read out of
// the repository being worked on:
//
// steps:
// - type: gate.diff-rules
// settings:
// rules:
// - id: no-new-panic
// level: error
// path: '\.go$'
// pattern: 'panic\('
// message: new panic() call -- use error returns instead
//
// The step diffs the change the branch has COMMITTED against the repository's
// base branch, matches each rule against the ADDED lines of that diff, and
// reports every match with its file and new-file line number. An error-level
// finding parks the run with the findings in the task's park reason; a
// warn-level finding is logged and nothing else. A rule that matches nothing is
// silent, and the number of added lines read is logged on every run, so
// "checked nothing" is distinguishable from "found nothing".
//
// Where the step sits is part of its contract, not a matter of taste: the rules
// read commits, never the worktree's own edits (Trees.Diff reports the merge
// base against HEAD), so the step must follow the step that commits the change.
// A run that would read no committed change while the worktree still holds
// uncommitted work fails on the spot and says which step it must follow,
// instead of reporting no findings for a change it never read.
//
// What it deliberately does not do is inspect a changed file's whole contents:
// a rule reads what the change adds, so a rule cannot be satisfied or broken by
// code the change did not touch. Anything a diff-shaped rule cannot express
// belongs in the repository's configured gate commands (config.Repo.Gate),
// which already run on the host.
const DiffRulesStepName = "gate.diff-rules"
// DiffRuleLevel values. An error-level finding parks the run; a warn-level
// finding is advisory and never blocks.
const (
DiffRuleLevelError = "error"
DiffRuleLevelWarn = "warn"
)
// diffRuleDetailBytes bounds how much of a park Detail the rendered findings
// occupy, matching the store's own park-reason cap headroom (see
// parkDetailBytes in prreview_own_pr.go).
const diffRuleDetailBytes = 4000
// DiffRule is one declarative rule applied to the added lines of the committed
// diff. Every field is an operator setting, so a rule is reviewable, versioned
// and audited with the definition it lives in -- none of which was true of the
// interpreted Go this replaces.
type DiffRule struct {
// ID names the rule in the run's log and in the park reason. Optional;
// without one a finding is still attributable to its file and line, not to
// the rule that reported it.
ID string `yaml:"id"`
// Level is DiffRuleLevelError or DiffRuleLevelWarn, and is required: a rule
// that does not say whether it blocks is refused rather than defaulted.
Level string `yaml:"level"`
// Pattern is the RE2 expression matched against each added line's text
// (without its leading "+"). Required.
Pattern string `yaml:"pattern"`
// Path, when set, is an RE2 expression the changed file's path must match
// for the rule to apply, e.g. '\.go$'. Optional.
Path string `yaml:"path"`
// ExcludePath, when set, is an RE2 expression the changed file's path must
// NOT match for the rule to apply, e.g. '_test\.go$'. Optional, and the
// only way to spare one class of file: RE2 has no lookahead.
ExcludePath string `yaml:"exclude_path"`
// Message is what the operator reads when the rule fires. Required.
Message string `yaml:"message"`
}
// diffRulesSettings is the step's whole settings surface. It exists as a type
// so the decode is strict: a settings key the step does not read is refused
// where the definition is validated, not ignored at run time.
type diffRulesSettings struct {
Rules []DiffRule `yaml:"rules"`
}
// DiffRulesStepType returns the registered step type for DiffRulesStepName, the
// way the shipped stages are declared: the factory builds the stage the
// executing side runs from the settings the validating side accepted.
func DiffRulesStepType() StepType {
return StepType{Name: DiffRulesStepName, Factory: newDiffRulesStage}
}
// newDiffRulesStage decodes and compiles the step's settings, refusing every
// contribution it cannot apply. This runs where the definition is validated and
// again where it is compiled, so a stored definition that cannot be honoured
// fails closed instead of loading and then quietly applying nothing.
func newDiffRulesStage(settings yaml.Node) (Stage, error) {
rules, err := decodeDiffRules(settings)
if err != nil {
return Stage{}, err
}
return Stage{Name: DiffRulesStepName, Run: func(ctx context.Context, tc *TaskContext) error {
return runDiffRules(ctx, tc, rules)
}}, nil
}
func runDiffRules(ctx context.Context, tc *TaskContext, rules []compiledDiffRule) error {
if tc.Dir == "" {
return fmt.Errorf("%s: no worktree to diff: this step must follow the one that prepares it", DiffRulesStepName)
}
base := tc.Repo.BaseBranch()
diff, err := tc.Trees.Diff(ctx, tc.Dir, base)
if err != nil {
return fmt.Errorf("%s: diff against %s: %w", DiffRulesStepName, base, err)
}
added := addedDiffLines(diff)
// The count is the run's own evidence of what the rules read: without it, a
// step that read nothing and a step whose rules matched nothing leave the
// same trace.
tc.Log.Info("diff rules read the committed change", "base", base, "added_lines", len(added), "rules", len(rules))
if diff == "" {
if err := refuseAChangeThatIsNotCommitted(ctx, tc); err != nil {
return err
}
}
var blocking []diffFinding
for _, finding := range matchDiffRules(rules, added) {
logRuleFinding(tc, finding)
if finding.Level == DiffRuleLevelError {
blocking = append(blocking, finding)
}
}
if len(blocking) > 0 {
// A blocking finding is a verdict about the change, not a failure of the
// run: the task parks with what was found, the way the adversarial
// review parks with its findings, so an operator reads why rather than
// an execution error that hides it.
tc.Outcome = Outcome{Status: StatusParked, Detail: renderDiffRuleFindings(blocking)}
}
return nil
}
// uncommittedChangeReporter is the optional capability through which a Trees
// implementation reports whether a worktree holds work no commit has captured
// yet: staged, unstaged or untracked. It is separate from the Trees contract for
// the same reason changeStatsReader is (steps.go): it is a capability one
// consumer needs, not a question every Trees implementation must answer. Both
// production implementations do answer it -- *worktree.Manager reads git, and
// hybridTrees forwards to its local manager, which is the path every production
// run takes.
type uncommittedChangeReporter interface {
HasUncommittedChanges(ctx context.Context, dir string) (bool, error)
}
// refuseAChangeThatIsNotCommitted fails the step when it read no committed
// change at all while the worktree still holds work. It is the guard the empty
// diff alone cannot provide: an empty diff is either "nothing has changed" or
// "the change is not committed yet", and only the worktree can say which.
//
// A clean worktree with nothing committed is a genuine no-op and is allowed. A
// Trees implementation that cannot answer at all is refused rather than assumed
// clean, because for a gate silence about the change is a failure, not a pass.
func refuseAChangeThatIsNotCommitted(ctx context.Context, tc *TaskContext) error {
const follow = "this step must follow the step that commits the change"
reporter, ok := tc.Trees.(uncommittedChangeReporter)
if !ok {
return fmt.Errorf("%s: no committed change to check and this worktree implementation cannot report uncommitted work: %s", DiffRulesStepName, follow)
}
uncommitted, err := reporter.HasUncommittedChanges(ctx, tc.Dir)
if err != nil {
return fmt.Errorf("%s: read uncommitted work: %w", DiffRulesStepName, err)
}
if !uncommitted {
return nil
}
return fmt.Errorf("%s: no committed change to check but the worktree holds uncommitted work: %s", DiffRulesStepName, follow)
}
// logRuleFinding records one finding under the severity of its effect -- a
// blocking finding is a warning an operator acts on, an advisory one is
// informational -- while the rule's own declared level rides along as
// rule_level. The field is not called "level": that key is slog's own record
// level, and a record carrying it twice is a coin toss for whichever reader
// keeps one.
func logRuleFinding(tc *TaskContext, finding diffFinding) {
attributes := []any{
"rule_level", finding.Level,
"rule", finding.Rule,
"file", finding.Path,
"line", finding.Line,
"message", finding.Message,
}
if finding.Level == DiffRuleLevelError {
tc.Log.Warn("diff rule finding", attributes...)
return
}
tc.Log.Info("diff rule finding", attributes...)
}
// renderDiffRuleFindings formats the blocking findings as a park reason: one
// line per violation, at the position a reader can jump to.
func renderDiffRuleFindings(findings []diffFinding) string {
var builder strings.Builder
builder.WriteString("diff rules found blocking violations:\n")
for _, finding := range findings {
fmt.Fprintf(&builder, "- %s:%d: %s", finding.Path, finding.Line, finding.Message)
if finding.Rule != "" {
fmt.Fprintf(&builder, " (%s)", finding.Rule)
}
builder.WriteString("\n")
}
return clip(builder.String(), diffRuleDetailBytes)
}
// decodeDiffRules strictly decodes the step's settings and compiles every rule
// up front.
func decodeDiffRules(settings yaml.Node) ([]compiledDiffRule, error) {
if settings.Kind == 0 {
return nil, fmt.Errorf("%s requires settings: a step with no rules would pass every change", DiffRulesStepName)
}
var decoded diffRulesSettings
if err := decodeSettingsStrict(settings, &decoded); err != nil {
return nil, fmt.Errorf("%s settings: %w", DiffRulesStepName, err)
}
if len(decoded.Rules) == 0 {
return nil, fmt.Errorf("%s: at least one rule is required", DiffRulesStepName)
}
rules := make([]compiledDiffRule, 0, len(decoded.Rules))
for i, rule := range decoded.Rules {
compiled, err := compileDiffRule(rule)
if err != nil {
return nil, fmt.Errorf("%s rule %d: %w", DiffRulesStepName, i+1, err)
}
rules = append(rules, compiled)
}
return rules, nil
}
// compileDiffRule turns one operator rule into the form the stage matches with,
// compiling its expressions here rather than at the first line of a run: a rule
// that cannot be applied is refused where the definition is validated.
func compileDiffRule(rule DiffRule) (compiledDiffRule, error) {
compiled := compiledDiffRule{id: rule.ID, level: rule.Level, message: rule.Message}
switch rule.Level {
case DiffRuleLevelError, DiffRuleLevelWarn:
case "":
return compiledDiffRule{}, errors.New("level is required (error or warn)")
default:
return compiledDiffRule{}, fmt.Errorf("unknown level %q, want %q or %q", rule.Level, DiffRuleLevelError, DiffRuleLevelWarn)
}
if rule.Message == "" {
return compiledDiffRule{}, errors.New("message is required")
}
if rule.Pattern == "" {
return compiledDiffRule{}, errors.New("pattern is required")
}
pattern, err := regexp.Compile(rule.Pattern)
if err != nil {
return compiledDiffRule{}, fmt.Errorf("pattern: %w", err)
}
compiled.pattern = pattern
if rule.Path != "" {
path, err := regexp.Compile(rule.Path)
if err != nil {
return compiledDiffRule{}, fmt.Errorf("path: %w", err)
}
compiled.path = path
}
if rule.ExcludePath != "" {
excludePath, err := regexp.Compile(rule.ExcludePath)
if err != nil {
return compiledDiffRule{}, fmt.Errorf("exclude_path: %w", err)
}
compiled.excludePath = excludePath
}
return compiled, nil
}
// decodeSettingsStrict decodes a settings node the way ParseDefinition's own
// decoder reads a definition: unknown fields are errors. yaml.Node.Decode is
// not strict, so the node is re-encoded and decoded with KnownFields rather
// than silently dropping a key this step does not read.
func decodeSettingsStrict(settings yaml.Node, target any) error {
raw, err := yaml.Marshal(&settings)
if err != nil {
return fmt.Errorf("re-encode settings: %w", err)
}
decoder := yaml.NewDecoder(bytes.NewReader(raw))
decoder.KnownFields(true)
return decoder.Decode(target)
}
// compiledDiffRule is one rule in the form the stage applies it: expressions
// compiled, level checked.
type compiledDiffRule struct {
id string
level string
message string
pattern *regexp.Regexp
path *regexp.Regexp
excludePath *regexp.Regexp
}
// appliesTo reports whether this rule covers a changed file's path.
func (r compiledDiffRule) appliesTo(path string) bool {
if r.path != nil && !r.path.MatchString(path) {
return false
}
return r.excludePath == nil || !r.excludePath.MatchString(path)
}
// diffFinding is one rule violation, at the position the change introduced it.
type diffFinding struct {
Rule string
Level string
Message string
Path string
Line int
}
// matchDiffRules applies every rule to every added line. Rules are not
// short-circuited by the first match: two rules can both object to one line,
// and an operator who wrote both wants to read both.
func matchDiffRules(rules []compiledDiffRule, added []diffLine) []diffFinding {
var findings []diffFinding
for _, line := range added {
for _, rule := range rules {
if !rule.appliesTo(line.Path) || !rule.pattern.MatchString(line.Text) {
continue
}
findings = append(findings, diffFinding{
Rule: rule.id,
Level: rule.level,
Message: rule.message,
Path: line.Path,
Line: line.Line,
})
}
}
return findings
}
// diffLine is one added line of a unified diff, positioned in the file the
// change produces -- the number an operator can open, not the diff's own
// offset.
type diffLine struct {
Path string
Line int
Text string
}
// addedDiffLines returns the lines a unified diff ADDS, each with the path of
// the file it lands in and its line number in the file the change produces.
//
// The diff is parsed by internal/domain/workflow/prreview, which owns the one
// parser this repository reads diffs with -- the review pipeline positions its
// findings by it. This maps that parser's result into the shape the rules
// matcher reads, so a rule and a review comment can never disagree about which
// line a change introduced.
func addedDiffLines(diff string) []diffLine {
added := prreview.AddedLines(prreview.ParseDiff(diff))
lines := make([]diffLine, 0, len(added))
for _, line := range added {
lines = append(lines, diffLine{Path: line.Path, Line: line.Line, Text: line.Text})
}
return lines
}