feat(store): user_identities map + UserBySubject/RegisterUser (ADR-012)

Add the Dex-subject -> tapir-user_id bridge for multi-user Stage 1. Migration 004 creates user_identities (dex_subject PK, user_id UNIQUE FK ON DELETE CASCADE). It is intentionally NOT RLS-enabled: it holds no user data and must be readable BEFORE a user_id is known (the lookup is what yields the id used to set tapir.current_user_id). RLS here would be a chicken-and-egg deadlock; data isolation stays on the user-owned tables. UserBySubject resolves subject -> user_id as a plain pool query (pre-scope, no withUser). RegisterUser generates the UUID app-side (stdlib crypto/rand, no new dep) so the forced-RLS WITH CHECK (id = GUC) passes, then inserts the users row via withUser(newID) and the identity row in the same transaction. Re-registration of a subject errors with ErrSubjectRegistered. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-03 15:53:34 +02:00
parent 9bff59037f
commit f396e01243
4 changed files with 217 additions and 0 deletions
@@ -0,0 +1,91 @@
+package store
+
+import (
+	"context"
+	"crypto/rand"
+	"errors"
+	"fmt"
+
+	"github.com/jackc/pgx/v5"
+)
+
+// ErrSubjectRegistered is returned by RegisterUser when the Dex subject already
+// maps to a tapir user. Registration is explicit and once-per-subject (ADR-012).
+var ErrSubjectRegistered = errors.New("store: subject already registered")
+
+// UserBySubject resolves a Dex subject to its tapir user_id via the un-RLS'd
+// user_identities map. It runs as a plain pool query WITHOUT withUser: this is
+// the pre-scope lookup whose result becomes the GUC for every subsequent
+// user-scoped access, so it cannot itself depend on that GUC being set. found is
+// false (no error) when the subject has no mapping yet — the caller routes such
+// requests to registration.
+func (s *Store) UserBySubject(ctx context.Context, subject string) (userID string, found bool, err error) {
+	err = s.pool.QueryRow(ctx,
+		`SELECT user_id FROM user_identities WHERE dex_subject = $1`, subject).Scan(&userID)
+	if errors.Is(err, pgx.ErrNoRows) {
+		return "", false, nil
+	}
+	if err != nil {
+		return "", false, fmt.Errorf("store: user by subject: %w", err)
+	}
+	return userID, true, nil
+}
+
+// RegisterUser creates the tapir user for a Dex subject and the identity mapping
+// that points to it, returning the new user_id. It errors with
+// ErrSubjectRegistered if the subject already maps.
+//
+// Bootstrapping note (generate-uuid-then-scope): the users table is FORCE'd RLS
+// with a WITH CHECK that defaults to the USING predicate id =
+// current_setting('tapir.current_user_id') (migration 003). A users row can
+// therefore only be inserted while the connection is ALREADY scoped to that
+// row's own id — a chicken-and-egg if the id were DB-generated. So we generate
+// the UUID app-side, scope to it via withUser(newID, ...), and insert the users
+// row inside that scope so the WITH CHECK passes. The user_identities row is
+// un-RLS'd auth plumbing; it is written in the SAME transaction so a user and
+// its mapping are always consistent.
+func (s *Store) RegisterUser(ctx context.Context, subject, displayName string) (userID string, err error) {
+	newID, err := newUUIDv4()
+	if err != nil {
+		return "", fmt.Errorf("store: register user: %w", err)
+	}
+
+	// Fast, clear rejection of a re-registration. The dex_subject PRIMARY KEY is
+	// the authoritative guard (a concurrent insert would still violate it); this
+	// check just turns the common case into a meaningful error instead of a raw
+	// constraint violation.
+	if _, found, err := s.UserBySubject(ctx, subject); err != nil {
+		return "", err
+	} else if found {
+		return "", fmt.Errorf("%w: %q", ErrSubjectRegistered, subject)
+	}
+
+	if err := s.withUser(ctx, newID, func(tx pgx.Tx) error {
+		if _, err := tx.Exec(ctx,
+			`INSERT INTO users (id, display_name) VALUES ($1, $2)`, newID, displayName); err != nil {
+			return fmt.Errorf("store: insert user: %w", err)
+		}
+		if _, err := tx.Exec(ctx,
+			`INSERT INTO user_identities (dex_subject, user_id) VALUES ($1, $2)`,
+			subject, newID); err != nil {
+			return fmt.Errorf("store: insert identity: %w", err)
+		}
+		return nil
+	}); err != nil {
+		return "", err
+	}
+	return newID, nil
+}
+
+// newUUIDv4 returns a random RFC-4122 v4 UUID string. Generated app-side (stdlib
+// crypto/rand, no new dependency) so the id is known before the row is scoped and
+// inserted — see RegisterUser's bootstrapping note.
+func newUUIDv4() (string, error) {
+	var b [16]byte
+	if _, err := rand.Read(b[:]); err != nil {
+		return "", fmt.Errorf("generate uuid: %w", err)
+	}
+	b[6] = (b[6] & 0x0f) | 0x40 // version 4
+	b[8] = (b[8] & 0x3f) | 0x80 // variant 10
+	return fmt.Sprintf("%x-%x-%x-%x-%x", b[0:4], b[4:6], b[6:8], b[8:10], b[10:16]), nil
+}
@@ -0,0 +1,103 @@
+package store_test
+
+import (
+	"context"
+	"errors"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+
+	"gitea.d-ma.be/mathias/tapir/internal/adapters/store"
+)
+
+const (
+	subjectA = "dex|alice-123"
+	subjectB = "dex|bob-456"
+)
+
+func TestUserBySubjectUnknownReturnsNotFound(t *testing.T) {
+	ctx := context.Background()
+	s := newStore(t)
+	resetDB(t, rawPool(t))
+
+	id, found, err := s.UserBySubject(ctx, subjectA)
+	require.NoError(t, err)
+	require.False(t, found)
+	require.Empty(t, id)
+}
+
+func TestRegisterUserCreatesUserAndIdentity(t *testing.T) {
+	ctx := context.Background()
+	s := newStore(t)
+	p := rawPool(t)
+	resetDB(t, p)
+
+	id, err := s.RegisterUser(ctx, subjectA, "Alice")
+	require.NoError(t, err)
+	require.NotEmpty(t, id)
+
+	// Exactly one users row with the returned id and the given display name.
+	var users int
+	var name string
+	require.NoError(t, p.QueryRow(ctx,
+		`SELECT count(*), coalesce(max(display_name), '') FROM users WHERE id = $1`, id).
+		Scan(&users, &name))
+	require.Equal(t, 1, users)
+	require.Equal(t, "Alice", name)
+
+	// Exactly one identity row mapping the subject to that id.
+	var idents int
+	require.NoError(t, p.QueryRow(ctx,
+		`SELECT count(*) FROM user_identities WHERE dex_subject = $1 AND user_id = $2`,
+		subjectA, id).Scan(&idents))
+	require.Equal(t, 1, idents)
+
+	// And it now resolves straight through.
+	got, found, err := s.UserBySubject(ctx, subjectA)
+	require.NoError(t, err)
+	require.True(t, found)
+	require.Equal(t, id, got)
+}
+
+func TestRegisterUserRejectsDuplicateSubject(t *testing.T) {
+	ctx := context.Background()
+	s := newStore(t)
+	p := rawPool(t)
+	resetDB(t, p)
+
+	first, err := s.RegisterUser(ctx, subjectA, "Alice")
+	require.NoError(t, err)
+
+	_, err = s.RegisterUser(ctx, subjectA, "Alice Again")
+	require.Error(t, err)
+	require.True(t, errors.Is(err, store.ErrSubjectRegistered))
+
+	// No second user was created; the original mapping is intact.
+	var users, idents int
+	require.NoError(t, p.QueryRow(ctx, `SELECT count(*) FROM users`).Scan(&users))
+	require.NoError(t, p.QueryRow(ctx, `SELECT count(*) FROM user_identities`).Scan(&idents))
+	require.Equal(t, 1, users)
+	require.Equal(t, 1, idents)
+
+	got, found, err := s.UserBySubject(ctx, subjectA)
+	require.NoError(t, err)
+	require.True(t, found)
+	require.Equal(t, first, got)
+}
+
+func TestRegisterUserDistinctSubjectsGetDistinctUsers(t *testing.T) {
+	ctx := context.Background()
+	s := newStore(t)
+	p := rawPool(t)
+	resetDB(t, p)
+
+	idA, err := s.RegisterUser(ctx, subjectA, "Alice")
+	require.NoError(t, err)
+	idB, err := s.RegisterUser(ctx, subjectB, "Bob")
+	require.NoError(t, err)
+	require.NotEqual(t, idA, idB)
+
+	var users int
+	require.NoError(t, p.QueryRow(ctx, `SELECT count(*) FROM users`).Scan(&users))
+	require.Equal(t, 2, users)
+}
@@ -0,0 +1 @@
+DROP TABLE IF EXISTS user_identities;
@@ -0,0 +1,22 @@
+-- Migration 004: the Dex-subject → tapir-user map (ADR-012 Stage 1, multi-user).
+-- A Dex-authenticated subject is the login identity; the tapir user_id (UUID) is
+-- what every user-owned, force-RLS table keys off. This table is the bridge:
+-- resolve subject → user_id here (auth plumbing, pre-scope), THEN scope all data
+-- access by that id via the store's withUser helper.
+--
+-- INTENTIONALLY NOT RLS-ENABLED. The forced-RLS isolation (migration 003) guards
+-- the user-OWNED data tables. user_identities holds no user data — only an opaque
+-- (dex_subject ↔ user_id) pair — and must be readable BEFORE a user_id is known
+-- (that lookup is what yields the id used to set tapir.current_user_id). Putting
+-- RLS here would be a chicken-and-egg deadlock (you'd need the GUC to read the row
+-- that tells you the GUC). Data isolation lives on the user-owned tables, not here.
+CREATE TABLE user_identities (
+    dex_subject TEXT PRIMARY KEY,
+    user_id     UUID NOT NULL UNIQUE REFERENCES users(id) ON DELETE CASCADE,
+    created_at  TIMESTAMPTZ NOT NULL DEFAULT now()
+);
+
+COMMENT ON TABLE user_identities IS
+    'Dex subject -> tapir user_id map. Auth plumbing, deliberately NOT RLS-enabled '
+    '(no user data; must be read pre-scope to resolve the id used for RLS). '
+    'ON DELETE CASCADE so deleting a user cleans up its identity mapping.';