docs/executing/provenance.rst (1)

23-30: Documentation lists fewer filesystem types than actually detected.

The implementation detects additional network filesystems (glusterfs, cifs, smbfs) that aren't mentioned here. Consider updating the parenthetical to include these, or use phrasing like "common network filesystems" to avoid maintaining an exhaustive list in documentation.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@docs/executing/provenance.rst` around lines 23 - 30, Update the parenthetical
in the opening sentence that lists detected network filesystems so it either
includes the additional types the implementation checks for (glusterfs, cifs,
smbfs) or replace the explicit list with a non-exhaustive phrase such as "common
network filesystems (e.g., NFS, CephFS, Lustre, GPFS)" to avoid divergence;
modify the text around the string starting "By default, snakemake automatically
detects if the database is located on a network filesystem" and keep the PRAGMA
examples unchanged.

src/snakemake/persistence/db.py (2)

260-271: The network filesystem type list may need extension.

As noted in the PR description, the detection list might need extension. Some commonly used parallel/network filesystems not currently included:

• beegfs (BeeGFS)
• afs (Andrew File System)
• pvfs2 / orangefs (Parallel Virtual File System)
• fuse.sshfs (SSHFS via FUSE)
• fhgfs (older BeeGFS name)

This is flagged for awareness rather than as a blocking issue, since false negatives simply result in more conservative (non-network) PRAGMA defaults.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/snakemake/persistence/db.py` around lines 260 - 271, The network_fs_types
set used for detecting network filesystems (the network_fs_types variable and
the membership check return fstype.casefold() in network_fs_types) is missing
additional common parallel/network FS names; update that set to include entries
like "beegfs", "fhgfs", "afs", "pvfs2", "orangefs", and "fuse.sshfs" (and
optionally "sshfs") so the fstype.casefold() membership test correctly detects
these filesystems.

---

247-258: Symlinks may cause incorrect filesystem detection.

Using os.path.abspath() doesn't resolve symbolic links. If the persistence path contains a symlink pointing to network storage (e.g., /home/user/data → /mnt/nfs/data), the detection will match against /home instead of /mnt/nfs, returning the wrong result.

Consider using os.path.realpath() to resolve symlinks before matching:

♻️ Proposed fix

-    path = os.path.abspath(path)
+    path = os.path.realpath(path)

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/snakemake/persistence/db.py` around lines 247 - 258, The code resolves
the persistence path with os.path.abspath() before matching mount points, which
doesn't follow symlinks and can cause incorrect filesystem detection; change the
resolution to use os.path.realpath() when assigning path so the matching loop
over psutil.disk_partitions(all=True) correctly compares the real mountpoint,
keeping the logic that updates longest_prefix_length and fstype unchanged.

✅ Autofix completed

src/snakemake/persistence/db.py (1)

83-95: ⚠️ Potential issue | 🟠 Major

Use engine.url.database instead of manually parsing the db_url string.

The current approach of manually stripping sqlite:/// and sqlite:// prefixes is brittle and fails for several valid SQLite URL formats. Driver-qualified URLs like sqlite+pysqlite:///..., URLs with query parameters, memory databases, and Windows paths all bypass this parsing logic. This causes is_network_filesystem() to inspect the wrong path, leading to incorrect PRAGMA selection.

SQLAlchemy provides the normalized database path via engine.url.database, which handles all valid SQLite URL formats correctly:

• sqlite:///relative.db → "relative.db"
• sqlite:////abs/path.db → "/abs/path.db"
• sqlite+pysqlite:////abs/path.db → "/abs/path.db"

Suggested change

-        # Parse SQLite file path from the DB URL for filesystem checks
-        sqlite_db_path = db_url
-        if sqlite_db_path.startswith("sqlite:///"):
-            sqlite_db_path = sqlite_db_path[len("sqlite:///"):]
-        elif sqlite_db_path.startswith("sqlite://"):
-            sqlite_db_path = sqlite_db_path[len("sqlite://"):]
-        # Remove query parameters and fragments if present
-        if "?" in sqlite_db_path:
-            sqlite_db_path = sqlite_db_path.split("?", 1)[0]
-        if "#" in sqlite_db_path:
-            sqlite_db_path = sqlite_db_path.split("#", 1)[0]
-
-        is_network_fs = is_network_filesystem(sqlite_db_path)
+        sqlite_db_path = (
+            self.engine.url.database if self.engine.dialect.name == "sqlite" else None
+        )
+        is_network_fs = bool(sqlite_db_path) and is_network_filesystem(sqlite_db_path)

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/snakemake/persistence/db.py` around lines 83 - 95, Replace the brittle
manual parsing of db_url used to derive sqlite_db_path with SQLAlchemy's
normalized value: read the database path from engine.url.database (use
engine.url.database instead of slicing db_url), then pass that to
is_network_filesystem and subsequent logic; update references to sqlite_db_path,
db_url, and is_network_fs accordingly so
is_network_filesystem(engine.url.database) is used and handles driver-qualified
URLs, memory DBs, query params, and Windows paths correctly.

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes