this repo has no description
lewis.moe
1use bytes::Bytes;
2use cid::Cid;
3use ipld_core::ipld::Ipld;
4use iroh_car::CarReader;
5use serde_json::Value as JsonValue;
6use sqlx::PgPool;
7use std::collections::HashMap;
8use std::io::Cursor;
9use thiserror::Error;
10use tracing::debug;
11use uuid::Uuid;
12
13#[derive(Error, Debug)]
14pub enum ImportError {
15 #[error("CAR parsing error: {0}")]
16 CarParse(String),
17 #[error("Expected exactly one root in CAR file")]
18 InvalidRootCount,
19 #[error("Block not found: {0}")]
20 BlockNotFound(String),
21 #[error("Invalid CBOR: {0}")]
22 InvalidCbor(String),
23 #[error("Database error: {0}")]
24 Database(#[from] sqlx::Error),
25 #[error("Block store error: {0}")]
26 BlockStore(String),
27 #[error("Import size limit exceeded")]
28 SizeLimitExceeded,
29 #[error("Repo not found")]
30 RepoNotFound,
31 #[error("Concurrent modification detected")]
32 ConcurrentModification,
33 #[error("Invalid commit structure: {0}")]
34 InvalidCommit(String),
35 #[error("Verification failed: {0}")]
36 VerificationFailed(#[from] super::verify::VerifyError),
37 #[error("DID mismatch: CAR is for {car_did}, but authenticated as {auth_did}")]
38 DidMismatch { car_did: String, auth_did: String },
39}
40
41#[derive(Debug, Clone)]
42pub struct BlobRef {
43 pub cid: String,
44 pub mime_type: Option<String>,
45}
46
47pub async fn parse_car(data: &[u8]) -> Result<(Cid, HashMap<Cid, Bytes>), ImportError> {
48 let cursor = Cursor::new(data);
49 let mut reader = CarReader::new(cursor)
50 .await
51 .map_err(|e| ImportError::CarParse(e.to_string()))?;
52 let header = reader.header();
53 let roots = header.roots();
54 if roots.len() != 1 {
55 return Err(ImportError::InvalidRootCount);
56 }
57 let root = roots[0];
58 let mut blocks = HashMap::new();
59 while let Ok(Some((cid, block))) = reader.next_block().await {
60 blocks.insert(cid, Bytes::from(block));
61 }
62 if !blocks.contains_key(&root) {
63 return Err(ImportError::BlockNotFound(root.to_string()));
64 }
65 Ok((root, blocks))
66}
67
68pub fn find_blob_refs_ipld(value: &Ipld, depth: usize) -> Vec<BlobRef> {
69 if depth > 32 {
70 return vec![];
71 }
72 match value {
73 Ipld::List(arr) => arr
74 .iter()
75 .flat_map(|v| find_blob_refs_ipld(v, depth + 1))
76 .collect(),
77 Ipld::Map(obj) => {
78 if let Some(Ipld::String(type_str)) = obj.get("$type")
79 && type_str == "blob"
80 && let Some(Ipld::Link(link_cid)) = obj.get("ref")
81 {
82 let mime = obj.get("mimeType").and_then(|v| {
83 if let Ipld::String(s) = v {
84 Some(s.clone())
85 } else {
86 None
87 }
88 });
89 return vec![BlobRef {
90 cid: link_cid.to_string(),
91 mime_type: mime,
92 }];
93 }
94 obj.values()
95 .flat_map(|v| find_blob_refs_ipld(v, depth + 1))
96 .collect()
97 }
98 _ => vec![],
99 }
100}
101
102pub fn find_blob_refs(value: &JsonValue, depth: usize) -> Vec<BlobRef> {
103 if depth > 32 {
104 return vec![];
105 }
106 match value {
107 JsonValue::Array(arr) => arr
108 .iter()
109 .flat_map(|v| find_blob_refs(v, depth + 1))
110 .collect(),
111 JsonValue::Object(obj) => {
112 if let Some(JsonValue::String(type_str)) = obj.get("$type")
113 && type_str == "blob"
114 && let Some(JsonValue::Object(ref_obj)) = obj.get("ref")
115 && let Some(JsonValue::String(link)) = ref_obj.get("$link")
116 {
117 let mime = obj
118 .get("mimeType")
119 .and_then(|v| v.as_str())
120 .map(String::from);
121 return vec![BlobRef {
122 cid: link.clone(),
123 mime_type: mime,
124 }];
125 }
126 obj.values()
127 .flat_map(|v| find_blob_refs(v, depth + 1))
128 .collect()
129 }
130 _ => vec![],
131 }
132}
133
134pub fn extract_links(value: &Ipld, links: &mut Vec<Cid>) {
135 match value {
136 Ipld::Link(cid) => {
137 links.push(*cid);
138 }
139 Ipld::Map(map) => {
140 for v in map.values() {
141 extract_links(v, links);
142 }
143 }
144 Ipld::List(arr) => {
145 for v in arr {
146 extract_links(v, links);
147 }
148 }
149 _ => {}
150 }
151}
152
153#[derive(Debug)]
154pub struct ImportedRecord {
155 pub collection: String,
156 pub rkey: String,
157 pub cid: Cid,
158 pub blob_refs: Vec<BlobRef>,
159}
160
161pub fn walk_mst(
162 blocks: &HashMap<Cid, Bytes>,
163 root_cid: &Cid,
164) -> Result<Vec<ImportedRecord>, ImportError> {
165 let mut records = Vec::new();
166 let mut stack = vec![*root_cid];
167 let mut visited = std::collections::HashSet::new();
168 while let Some(cid) = stack.pop() {
169 if visited.contains(&cid) {
170 continue;
171 }
172 visited.insert(cid);
173 let block = blocks
174 .get(&cid)
175 .ok_or_else(|| ImportError::BlockNotFound(cid.to_string()))?;
176 let value: Ipld = serde_ipld_dagcbor::from_slice(block)
177 .map_err(|e| ImportError::InvalidCbor(e.to_string()))?;
178 if let Ipld::Map(ref obj) = value {
179 if let Some(Ipld::List(entries)) = obj.get("e") {
180 for entry in entries {
181 if let Ipld::Map(entry_obj) = entry {
182 let key = entry_obj.get("k").and_then(|k| {
183 if let Ipld::Bytes(b) = k {
184 String::from_utf8(b.clone()).ok()
185 } else if let Ipld::String(s) = k {
186 Some(s.clone())
187 } else {
188 None
189 }
190 });
191 let record_cid = entry_obj.get("v").and_then(|v| {
192 if let Ipld::Link(cid) = v {
193 Some(*cid)
194 } else {
195 None
196 }
197 });
198 if let (Some(key), Some(record_cid)) = (key, record_cid)
199 && let Some(record_block) = blocks.get(&record_cid)
200 && let Ok(record_value) =
201 serde_ipld_dagcbor::from_slice::<Ipld>(record_block)
202 {
203 let blob_refs = find_blob_refs_ipld(&record_value, 0);
204 let parts: Vec<&str> = key.split('/').collect();
205 if parts.len() >= 2 {
206 let collection = parts[..parts.len() - 1].join("/");
207 let rkey = parts[parts.len() - 1].to_string();
208 records.push(ImportedRecord {
209 collection,
210 rkey,
211 cid: record_cid,
212 blob_refs,
213 });
214 }
215 }
216 if let Some(Ipld::Link(tree_cid)) = entry_obj.get("t") {
217 stack.push(*tree_cid);
218 }
219 }
220 }
221 }
222 if let Some(Ipld::Link(left_cid)) = obj.get("l") {
223 stack.push(*left_cid);
224 }
225 }
226 }
227 Ok(records)
228}
229
230pub struct CommitInfo {
231 pub rev: Option<String>,
232 pub prev: Option<String>,
233}
234
235fn extract_commit_info(commit: &Ipld) -> Result<(Cid, CommitInfo), ImportError> {
236 let obj = match commit {
237 Ipld::Map(m) => m,
238 _ => {
239 return Err(ImportError::InvalidCommit(
240 "Commit must be a map".to_string(),
241 ));
242 }
243 };
244 let data_cid = obj
245 .get("data")
246 .and_then(|d| {
247 if let Ipld::Link(cid) = d {
248 Some(*cid)
249 } else {
250 None
251 }
252 })
253 .ok_or_else(|| ImportError::InvalidCommit("Missing data field".to_string()))?;
254 let rev = obj.get("rev").and_then(|r| {
255 if let Ipld::String(s) = r {
256 Some(s.clone())
257 } else {
258 None
259 }
260 });
261 let prev = obj.get("prev").and_then(|p| {
262 if let Ipld::Link(cid) = p {
263 Some(cid.to_string())
264 } else if let Ipld::Null = p {
265 None
266 } else {
267 None
268 }
269 });
270 Ok((data_cid, CommitInfo { rev, prev }))
271}
272
273pub async fn apply_import(
274 db: &PgPool,
275 user_id: Uuid,
276 root: Cid,
277 blocks: HashMap<Cid, Bytes>,
278 max_blocks: usize,
279) -> Result<Vec<ImportedRecord>, ImportError> {
280 if blocks.len() > max_blocks {
281 return Err(ImportError::SizeLimitExceeded);
282 }
283 let root_block = blocks
284 .get(&root)
285 .ok_or_else(|| ImportError::BlockNotFound(root.to_string()))?;
286 let commit: Ipld = serde_ipld_dagcbor::from_slice(root_block)
287 .map_err(|e| ImportError::InvalidCbor(e.to_string()))?;
288 let (data_cid, _commit_info) = extract_commit_info(&commit)?;
289 let records = walk_mst(&blocks, &data_cid)?;
290 debug!(
291 "Importing {} blocks and {} records for user {}",
292 blocks.len(),
293 records.len(),
294 user_id
295 );
296 let mut tx = db.begin().await?;
297 let repo = sqlx::query!(
298 "SELECT repo_root_cid FROM repos WHERE user_id = $1 FOR UPDATE NOWAIT",
299 user_id
300 )
301 .fetch_optional(&mut *tx)
302 .await
303 .map_err(|e| {
304 if let sqlx::Error::Database(ref db_err) = e
305 && db_err.code().as_deref() == Some("55P03")
306 {
307 return ImportError::ConcurrentModification;
308 }
309 ImportError::Database(e)
310 })?;
311 if repo.is_none() {
312 return Err(ImportError::RepoNotFound);
313 }
314 let block_chunks: Vec<Vec<(&Cid, &Bytes)>> = blocks
315 .iter()
316 .collect::<Vec<_>>()
317 .chunks(100)
318 .map(|c| c.to_vec())
319 .collect();
320 for chunk in block_chunks {
321 for (cid, data) in chunk {
322 let cid_bytes = cid.to_bytes();
323 sqlx::query!(
324 "INSERT INTO blocks (cid, data) VALUES ($1, $2) ON CONFLICT (cid) DO NOTHING",
325 &cid_bytes,
326 data.as_ref()
327 )
328 .execute(&mut *tx)
329 .await?;
330 }
331 }
332 let root_str = root.to_string();
333 sqlx::query!(
334 "UPDATE repos SET repo_root_cid = $1, updated_at = NOW() WHERE user_id = $2",
335 root_str,
336 user_id
337 )
338 .execute(&mut *tx)
339 .await?;
340 sqlx::query!("DELETE FROM records WHERE repo_id = $1", user_id)
341 .execute(&mut *tx)
342 .await?;
343 for record in &records {
344 let record_cid_str = record.cid.to_string();
345 sqlx::query!(
346 r#"
347 INSERT INTO records (repo_id, collection, rkey, record_cid)
348 VALUES ($1, $2, $3, $4)
349 ON CONFLICT (repo_id, collection, rkey) DO UPDATE SET record_cid = $4
350 "#,
351 user_id,
352 record.collection,
353 record.rkey,
354 record_cid_str
355 )
356 .execute(&mut *tx)
357 .await?;
358 }
359 tx.commit().await?;
360 debug!(
361 "Successfully imported {} blocks and {} records",
362 blocks.len(),
363 records.len()
364 );
365 Ok(records)
366}
367
368#[cfg(test)]
369mod tests {
370 use super::*;
371
372 #[test]
373 fn test_find_blob_refs() {
374 let record = serde_json::json!({
375 "$type": "app.bsky.feed.post",
376 "text": "Hello world",
377 "embed": {
378 "$type": "app.bsky.embed.images",
379 "images": [
380 {
381 "alt": "Test image",
382 "image": {
383 "$type": "blob",
384 "ref": {
385 "$link": "bafkreihdwdcefgh4dqkjv67uzcmw7ojee6xedzdetojuzjevtenxquvyku"
386 },
387 "mimeType": "image/jpeg",
388 "size": 12345
389 }
390 }
391 ]
392 }
393 });
394 let blob_refs = find_blob_refs(&record, 0);
395 assert_eq!(blob_refs.len(), 1);
396 assert_eq!(
397 blob_refs[0].cid,
398 "bafkreihdwdcefgh4dqkjv67uzcmw7ojee6xedzdetojuzjevtenxquvyku"
399 );
400 assert_eq!(blob_refs[0].mime_type, Some("image/jpeg".to_string()));
401 }
402
403 #[test]
404 fn test_find_blob_refs_no_blobs() {
405 let record = serde_json::json!({
406 "$type": "app.bsky.feed.post",
407 "text": "Hello world"
408 });
409 let blob_refs = find_blob_refs(&record, 0);
410 assert!(blob_refs.is_empty());
411 }
412
413 #[test]
414 fn test_find_blob_refs_depth_limit() {
415 fn deeply_nested(depth: usize) -> JsonValue {
416 if depth == 0 {
417 serde_json::json!({
418 "$type": "blob",
419 "ref": { "$link": "bafkreitest" },
420 "mimeType": "image/png"
421 })
422 } else {
423 serde_json::json!({ "nested": deeply_nested(depth - 1) })
424 }
425 }
426 let deep = deeply_nested(40);
427 let blob_refs = find_blob_refs(&deep, 0);
428 assert!(blob_refs.is_empty());
429 }
430}