assay-auth 0.6.0

Authentication, OIDC (client + provider), passkey, Argon2, JWT, Biscuit capability tokens, session management, and Zanzibar-style authorization for assay-engine.
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
//! SQLite [`ZanzibarStore`] implementation — same recursive-CTE shape
//! as the Postgres backend, with the dialect differences absorbed:
//!
//! - `JSONB` → `TEXT`. We round-trip through `serde_json::to_string` /
//!   `from_str` rather than relying on `sqlx::types::Json` so the
//!   storage format is human-readable in the file (matters when an
//!   operator opens `data/auth.db` to debug).
//! - `text[]` → JSON-encoded array consumed by `json_each`. SQLite
//!   doesn't have a native array type; the standard idiom is `WHERE
//!   relation IN (SELECT value FROM json_each(?))`.
//! - `IS NOT DISTINCT FROM` doesn't exist; SQLite's `IS` operator
//!   already treats NULL=NULL as true, so direct-tuple delete uses
//!   `subject_rel IS ?`.
//! - The cycle-guard "path" is encoded as a JSON array via
//!   `json_array(...)` + `json_array_length` for the depth limit, and
//!   `instr(json_path, key)` for the membership check — matches the
//!   PG `path || ROW(...) ANY` semantic in spirit.

use std::sync::Arc;

use anyhow::{Context, Result};
use sqlx::{Row, SqlitePool};

use super::eval::{LeafCheck, Verdict, evaluate};
use super::resolve::resolve;
use super::store::ZanzibarStore;
use super::types::{
    CheckResult, Consistency, MAX_DEPTH, NamespaceSchema, ObjectRef, SubjectRef, TreeOp, Tuple,
    TupleFilter, UsersetTree,
};
use std::future::Future;
use std::pin::Pin;

/// SQLite-backed Zanzibar store. Cheap to clone (the underlying pool
/// is `Arc`d).
#[derive(Clone)]
pub struct SqliteZanzibarStore {
    pool: SqlitePool,
}

impl SqliteZanzibarStore {
    pub fn new(pool: SqlitePool) -> Self {
        Self { pool }
    }

    pub fn into_dyn(self) -> Arc<dyn ZanzibarStore> {
        Arc::new(self)
    }
}

fn now_secs() -> f64 {
    std::time::SystemTime::now()
        .duration_since(std::time::UNIX_EPOCH)
        .unwrap_or_default()
        .as_secs_f64()
}

#[async_trait::async_trait]
impl ZanzibarStore for SqliteZanzibarStore {
    async fn define_namespace(&self, schema: &NamespaceSchema) -> Result<()> {
        let json = serde_json::to_string(schema).context("zanzibar serialize NamespaceSchema")?;
        let now = now_secs();
        sqlx::query(
            "INSERT INTO auth.zanzibar_namespaces (name, schema_json, updated_at)
             VALUES (?, ?, ?)
             ON CONFLICT (name) DO UPDATE
                 SET schema_json = excluded.schema_json,
                     updated_at  = excluded.updated_at",
        )
        .bind(&schema.name)
        .bind(json)
        .bind(now)
        .execute(&self.pool)
        .await
        .context("auth.zanzibar_namespaces upsert")?;
        Ok(())
    }

    async fn get_namespace(&self, name: &str) -> Result<Option<NamespaceSchema>> {
        let row: Option<(String,)> =
            sqlx::query_as("SELECT schema_json FROM auth.zanzibar_namespaces WHERE name = ?")
                .bind(name)
                .fetch_optional(&self.pool)
                .await
                .context("auth.zanzibar_namespaces get")?;
        Ok(match row {
            Some((json,)) => {
                Some(serde_json::from_str(&json).context("zanzibar deserialize NamespaceSchema")?)
            }
            None => None,
        })
    }

    async fn list_namespaces(&self) -> Result<Vec<NamespaceSchema>> {
        let rows: Vec<(String,)> =
            sqlx::query_as("SELECT schema_json FROM auth.zanzibar_namespaces ORDER BY name")
                .fetch_all(&self.pool)
                .await
                .context("auth.zanzibar_namespaces list")?;
        rows.into_iter()
            .map(|(json,)| {
                serde_json::from_str(&json).context("zanzibar deserialize NamespaceSchema")
            })
            .collect()
    }

    async fn write_tuple(&self, t: &Tuple) -> Result<()> {
        let now = now_secs();
        sqlx::query(
            "INSERT INTO auth.zanzibar_tuples
                (object_type, object_id, relation,
                 subject_type, subject_id, subject_rel, created_at)
             VALUES (?, ?, ?, ?, ?, ?, ?)
             ON CONFLICT DO NOTHING",
        )
        .bind(&t.object_type)
        .bind(&t.object_id)
        .bind(&t.relation)
        .bind(&t.subject_type)
        .bind(&t.subject_id)
        .bind(&t.subject_rel)
        .bind(now)
        .execute(&self.pool)
        .await
        .context("auth.zanzibar_tuples insert")?;
        Ok(())
    }

    async fn write_tuples(&self, tuples: &[Tuple]) -> Result<()> {
        if tuples.is_empty() {
            return Ok(());
        }
        let mut tx = self.pool.begin().await.context("begin tuples txn")?;
        for t in tuples {
            let now = now_secs();
            sqlx::query(
                "INSERT INTO auth.zanzibar_tuples
                    (object_type, object_id, relation,
                     subject_type, subject_id, subject_rel, created_at)
                 VALUES (?, ?, ?, ?, ?, ?, ?)
                 ON CONFLICT DO NOTHING",
            )
            .bind(&t.object_type)
            .bind(&t.object_id)
            .bind(&t.relation)
            .bind(&t.subject_type)
            .bind(&t.subject_id)
            .bind(&t.subject_rel)
            .bind(now)
            .execute(&mut *tx)
            .await
            .context("auth.zanzibar_tuples batch insert")?;
        }
        tx.commit().await.context("commit tuples txn")?;
        Ok(())
    }

    async fn delete_tuple(&self, t: &Tuple) -> Result<bool> {
        // subject_rel is NOT NULL ('' for direct), so plain equality
        // works on both backends — no `IS` / `IS NOT DISTINCT FROM`
        // dance needed.
        let res = sqlx::query(
            "DELETE FROM auth.zanzibar_tuples
             WHERE object_type = ? AND object_id = ? AND relation = ?
               AND subject_type = ? AND subject_id = ?
               AND subject_rel = ?",
        )
        .bind(&t.object_type)
        .bind(&t.object_id)
        .bind(&t.relation)
        .bind(&t.subject_type)
        .bind(&t.subject_id)
        .bind(&t.subject_rel)
        .execute(&self.pool)
        .await
        .context("auth.zanzibar_tuples delete")?;
        Ok(res.rows_affected() > 0)
    }

    async fn list_tuples(&self, filter: &TupleFilter) -> Result<Vec<Tuple>> {
        // Each filter field uses the `(? IS NULL OR col = ?)` idiom so a
        // single static SQL string covers every filter combination.
        // sqlx doesn't reuse `?N`-style placeholders across the two
        // backends, so we bind the same Option twice per field.
        let rows: Vec<(String, String, String, String, String, String)> = sqlx::query_as(
            "SELECT object_type, object_id, relation,
                    subject_type, subject_id, subject_rel
             FROM auth.zanzibar_tuples
             WHERE (? IS NULL OR object_type  = ?)
               AND (? IS NULL OR object_id    = ?)
               AND (? IS NULL OR relation     = ?)
               AND (? IS NULL OR subject_type = ?)
               AND (? IS NULL OR subject_id   = ?)
             ORDER BY object_type, object_id, relation, subject_type, subject_id
             LIMIT ? OFFSET ?",
        )
        .bind(&filter.object_type)
        .bind(&filter.object_type)
        .bind(&filter.object_id)
        .bind(&filter.object_id)
        .bind(&filter.relation)
        .bind(&filter.relation)
        .bind(&filter.subject_type)
        .bind(&filter.subject_type)
        .bind(&filter.subject_id)
        .bind(&filter.subject_id)
        .bind(filter.effective_limit())
        .bind(filter.effective_offset())
        .fetch_all(&self.pool)
        .await
        .context("auth.zanzibar_tuples list")?;
        Ok(rows
            .into_iter()
            .map(|(ot, oid, rel, st, sid, srel)| Tuple {
                object_type: ot,
                object_id: oid,
                relation: rel,
                subject_type: st,
                subject_id: sid,
                subject_rel: srel,
            })
            .collect())
    }

    async fn check(
        &self,
        resource: &ObjectRef,
        permission: &str,
        subject: &SubjectRef,
        _consistency: Consistency,
    ) -> Result<CheckResult> {
        // No namespace defined → deny (the safe default). The full
        // permission algebra (union / intersect / exclude / arrow) is
        // composed by the backend-agnostic evaluator over the two
        // primitives this backend exposes (`check_relation_set` +
        // `arrow_targets`); a permission that flattens to a pure union
        // of relations still hits the single-CTE fast path inside it.
        let Some(schema) = self.get_namespace(&resource.object_type).await? else {
            return Ok(CheckResult::Denied);
        };
        evaluate(self, &schema, resource, permission, subject).await
    }

    async fn expand(
        &self,
        resource: &ObjectRef,
        relation: &str,
        depth_limit: u32,
    ) -> Result<UsersetTree> {
        let depth = depth_limit.min(MAX_DEPTH);
        Ok(UsersetTree::Node {
            op: TreeOp::Direct,
            children: expand_sqlite(self, resource, relation, depth, &mut Vec::new()).await?,
        })
    }

    async fn lookup_resources(
        &self,
        resource_type: &str,
        permission: &str,
        subject: &SubjectRef,
    ) -> Result<Vec<ObjectRef>> {
        let Some(schema) = self.get_namespace(resource_type).await? else {
            return Ok(Vec::new());
        };
        let Some(resolved) = resolve(&schema, permission) else {
            return Ok(Vec::new());
        };
        if resolved.union_relations.is_empty() {
            return Ok(Vec::new());
        }
        let relation_list: Vec<String> = resolved.union_relations.into_iter().collect();
        let relation_json =
            serde_json::to_string(&relation_list).context("encode relation list as JSON")?;
        let rows = sqlx::query(
            "SELECT DISTINCT object_type, object_id
             FROM auth.zanzibar_tuples
             WHERE object_type = ?
               AND relation IN (SELECT value FROM json_each(?))",
        )
        .bind(resource_type)
        .bind(relation_json)
        .fetch_all(&self.pool)
        .await
        .context("auth.zanzibar_tuples candidate resources")?;
        let mut out = Vec::new();
        for row in rows {
            let object_type: String = row.get("object_type");
            let object_id: String = row.get("object_id");
            let r = ObjectRef::new(object_type, object_id);
            if self
                .check(&r, permission, subject, Consistency::Minimum)
                .await?
                .is_allowed()
            {
                out.push(r);
            }
        }
        Ok(out)
    }

    async fn lookup_subjects(
        &self,
        subject_type: &str,
        resource: &ObjectRef,
        permission: &str,
    ) -> Result<Vec<SubjectRef>> {
        let Some(schema) = self.get_namespace(&resource.object_type).await? else {
            return Ok(Vec::new());
        };
        let Some(resolved) = resolve(&schema, permission) else {
            return Ok(Vec::new());
        };
        if resolved.union_relations.is_empty() {
            return Ok(Vec::new());
        }
        let relation_list: Vec<String> = resolved.union_relations.into_iter().collect();
        let relation_json =
            serde_json::to_string(&relation_list).context("encode relation list as JSON")?;

        let rows = sqlx::query(
            r#"
            WITH RECURSIVE walk(subject_type, subject_id, subject_rel, depth, path) AS (
                SELECT t.subject_type, t.subject_id, t.subject_rel, 1,
                       '|' || t.subject_type || ':' || t.subject_id || '|'
                FROM auth.zanzibar_tuples t
                WHERE t.object_type = ?1
                  AND t.object_id   = ?2
                  AND t.relation IN (SELECT value FROM json_each(?3))
                UNION ALL
                SELECT t.subject_type, t.subject_id, t.subject_rel, w.depth + 1,
                       w.path || t.subject_type || ':' || t.subject_id || '|'
                FROM auth.zanzibar_tuples t
                JOIN walk w
                  ON t.object_type = w.subject_type
                 AND t.object_id   = w.subject_id
                 AND w.subject_rel <> ''
                 AND t.relation = w.subject_rel
                WHERE w.depth < ?4
                  AND instr(w.path, '|' || t.subject_type || ':' || t.subject_id || '|') = 0
            )
            SELECT DISTINCT subject_type, subject_id
            FROM walk
            WHERE subject_type = ?5 AND subject_rel = ''
            "#,
        )
        .bind(&resource.object_type)
        .bind(&resource.object_id)
        .bind(relation_json)
        .bind(MAX_DEPTH as i64)
        .bind(subject_type)
        .fetch_all(&self.pool)
        .await
        .context("auth.zanzibar lookup_subjects CTE")?;
        Ok(rows
            .into_iter()
            .map(|row| {
                SubjectRef::direct(
                    row.get::<String, _>("subject_type"),
                    row.get::<String, _>("subject_id"),
                )
            })
            .collect())
    }
}

impl LeafCheck for SqliteZanzibarStore {
    fn check_relation_set<'a>(
        &'a self,
        object: &'a ObjectRef,
        relations: &'a [String],
        subject: &'a SubjectRef,
    ) -> Pin<Box<dyn Future<Output = Result<Verdict>> + Send + 'a>> {
        Box::pin(async move {
            if relations.is_empty() {
                return Ok(Verdict::Denied);
            }
            // SQLite has no array type — encode the relation set as a
            // JSON array consumed via `json_each(?)`. Same shape as PG's
            // `text[]` parameter, just one extra serialisation step.
            let relation_json =
                serde_json::to_string(relations).context("encode relation list as JSON")?;

            // Cycle-guard via a JSON-encoded path string. Each hop
            // appends `<type>:<id>` and the join refuses to add a node
            // already on the path. SQLite recursive CTEs allow this with
            // ordinary `||` string concat.
            let row: Option<(i64,)> = sqlx::query_as(
                r#"
                WITH RECURSIVE walk(subject_type, subject_id, subject_rel, depth, path) AS (
                    SELECT t.subject_type,
                           t.subject_id,
                           t.subject_rel,
                           1 AS depth,
                           '|' || t.subject_type || ':' || t.subject_id || '|' AS path
                    FROM auth.zanzibar_tuples t
                    WHERE t.object_type = ?1
                      AND t.object_id = ?2
                      AND t.relation IN (SELECT value FROM json_each(?3))
                    UNION ALL
                    SELECT t.subject_type,
                           t.subject_id,
                           t.subject_rel,
                           w.depth + 1,
                           w.path || t.subject_type || ':' || t.subject_id || '|'
                    FROM auth.zanzibar_tuples t
                    JOIN walk w
                      ON t.object_type = w.subject_type
                     AND t.object_id   = w.subject_id
                     AND w.subject_rel <> ''
                     AND t.relation = w.subject_rel
                    WHERE w.depth < ?4
                      AND instr(w.path, '|' || t.subject_type || ':' || t.subject_id || '|') = 0
                )
                SELECT CASE
                    WHEN EXISTS (
                        SELECT 1 FROM walk
                        WHERE subject_type = ?5 AND subject_id = ?6 AND subject_rel = ''
                    ) THEN 1
                    WHEN EXISTS (SELECT 1 FROM walk WHERE depth >= ?4) THEN 2
                    ELSE 0
                END AS verdict
                "#,
            )
            .bind(&object.object_type)
            .bind(&object.object_id)
            .bind(relation_json)
            .bind(MAX_DEPTH as i64)
            .bind(&subject.subject_type)
            .bind(&subject.subject_id)
            .fetch_optional(&self.pool)
            .await
            .context("auth.zanzibar check CTE")?;

            Ok(match row.map(|(v,)| v).unwrap_or(0) {
                1 => Verdict::Allowed,
                2 => Verdict::DepthExceeded,
                _ => Verdict::Denied,
            })
        })
    }

    fn arrow_targets<'a>(
        &'a self,
        object: &'a ObjectRef,
        relation: &'a str,
    ) -> Pin<Box<dyn Future<Output = Result<Vec<ObjectRef>>> + Send + 'a>> {
        Box::pin(async move {
            // Only object-reference subjects (`subject_rel = ''`) are
            // the left edge of an arrow hop — a userset subject on the
            // tupleset relation has no object to re-evaluate against.
            let rows = sqlx::query(
                "SELECT subject_type, subject_id
                 FROM auth.zanzibar_tuples
                 WHERE object_type = ? AND object_id = ? AND relation = ?
                   AND subject_rel = ''",
            )
            .bind(&object.object_type)
            .bind(&object.object_id)
            .bind(relation)
            .fetch_all(&self.pool)
            .await
            .context("auth.zanzibar arrow targets")?;
            Ok(rows
                .into_iter()
                .map(|row| {
                    ObjectRef::new(
                        row.get::<String, _>("subject_type"),
                        row.get::<String, _>("subject_id"),
                    )
                })
                .collect())
        })
    }

    fn schema_for<'a>(
        &'a self,
        object_type: &'a str,
    ) -> Pin<Box<dyn Future<Output = Result<Option<NamespaceSchema>>> + Send + 'a>> {
        Box::pin(async move { self.get_namespace(object_type).await })
    }
}

/// Recursive helper for `expand` — same shape as the PG version, just
/// SQLite's parameter binding (`?` instead of `$N`).
fn expand_sqlite<'a>(
    store: &'a SqliteZanzibarStore,
    resource: &'a ObjectRef,
    relation: &'a str,
    depth: u32,
    seen: &'a mut Vec<String>,
) -> std::pin::Pin<Box<dyn std::future::Future<Output = Result<Vec<UsersetTree>>> + Send + 'a>> {
    Box::pin(async move {
        if depth == 0 {
            return Ok(Vec::new());
        }
        let key = format!(
            "{}:{}#{}",
            resource.object_type, resource.object_id, relation
        );
        if seen.contains(&key) {
            return Ok(Vec::new());
        }
        seen.push(key);
        let rows = sqlx::query(
            "SELECT subject_type, subject_id, subject_rel
             FROM auth.zanzibar_tuples
             WHERE object_type = ? AND object_id = ? AND relation = ?",
        )
        .bind(&resource.object_type)
        .bind(&resource.object_id)
        .bind(relation)
        .fetch_all(&store.pool)
        .await
        .context("auth.zanzibar_tuples expand fetch")?;
        let mut children = Vec::new();
        for row in rows {
            let st: String = row.get("subject_type");
            let sid: String = row.get("subject_id");
            let sr: String = row.get("subject_rel");
            if sr.is_empty() {
                children.push(UsersetTree::Leaf {
                    subject: SubjectRef::direct(st, sid),
                });
            } else {
                let inner_resource = ObjectRef::new(st.clone(), sid.clone());
                let sub = expand_sqlite(store, &inner_resource, &sr, depth - 1, seen).await?;
                children.push(UsersetTree::Node {
                    op: TreeOp::TuplesetArrow,
                    children: vec![
                        UsersetTree::Leaf {
                            subject: SubjectRef::userset(st, sid, sr.clone()),
                        },
                        UsersetTree::Node {
                            op: TreeOp::Direct,
                            children: sub,
                        },
                    ],
                });
            }
        }
        Ok(children)
    })
}