Merge pull request #24 from snegov/feature/parse-library-list-as-json

Read the local library list as JSON instead of splitting columns

Dockerfile

@@ -68,6 +68,8 @@ RUN chmod +x /dsc/start.py && \
 # Smoke checks: seaf-cli has no --version flag, so assert its subcommands, and
 # confirm the app's own imports resolve.
 RUN seaf-cli --help | grep -q 'download-by-name' && \
+    seaf-cli sync --help | grep -q '\--token' && \
+    seaf-cli list --help | grep -q '\--json' && \
     python3 -c "import seafile, pysearpc, dsc.client, dsc.misc"
 
 VOLUME /dsc/seafile-data

IMPROVEMENTS.md

@@ -152,7 +152,10 @@ Acceptance criteria:
   server errors separately.
 - [ ] Make TLS configuration consistent between Python requests and Seafile.
 - [ ] Support a custom CA bundle; keep certificate disabling as a last resort.
-- [ ] Prefer structured JSON output from `seaf-cli` over parsing display text.
+- [x] Prefer structured JSON output from `seaf-cli` over parsing display text.
+  `seaf-cli list --json` is used instead of splitting the display columns,
+  which could not represent a library name or a path containing whitespace.
+  `list-remote` still uses the HTTP API rather than the CLI.
 - [ ] Reject ambiguous library names and an empty requested library set.
 
 ### Runtime hardening

dsc/client.py

@@ -1,4 +1,5 @@
 import argparse
+import json
 import logging
 import os
 import subprocess
@@ -271,16 +272,27 @@ class SeafileClient:
                 prev_status[library] = cur_status[library]
 
     def get_local_libraries(self) -> set:
-        cmd = ["seaf-cli", "list"]
+        # The display output is whitespace separated, so a library name or a
+        # path containing a space cannot be parsed back apart. Ask for JSON.
+        cmd = ["seaf-cli", "list", "--json"]
         _lg.info("Listing local libraries: %s", " ".join(cmd))
-        out = subprocess.check_output(self.__gen_cmd(cmd))
-        out = out.decode().splitlines()[1:]  # first line is a header
+        out = self.__run(cmd, stdout=subprocess.PIPE).stdout
 
-        local_libs = set()
-        for line in out:
-            lib_name, lib_id, lib_path = line.rsplit(maxsplit=3)
-            local_libs.add(lib_id)
-        return local_libs
+        try:
+            libraries = json.loads(out.decode())
+        except (UnicodeDecodeError, ValueError) as err:
+            raise DaemonError(
+                f"Cannot read the library list from seaf-cli: {err}"
+            ) from err
+
+        # Valid JSON is not necessarily the expected shape, and an unexpected
+        # one must fail the same way rather than as a bare TypeError.
+        try:
+            return {lib["id"] for lib in libraries}
+        except (KeyError, TypeError) as err:
+            raise DaemonError(
+                f"Unexpected library list from seaf-cli: {err}"
+            ) from err
 
     def configure(self, args: argparse.Namespace, check_for_daemon: bool = True):
         need_restart = False

tests/test_commands.py

@@ -1,12 +1,14 @@
 """Every subprocess must be an argument list, never a shell string."""
 
 import argparse
+import json
 import subprocess
 
 import pytest
 
 from dsc import client as client_module
 from dsc import const
+from dsc.errors import DaemonError
 from dsc.client import SeafileClient
 from dsc.misc import hide_secrets, user_cmd
 
@@ -157,3 +159,79 @@ def test_configure_passes_option_values_as_arguments(client, calls, monkeypatch)
     assert prefix + ["-k", "upload_limit", "-v", "100"] in argvs
     # An option that seaf-cli does not know about is never forwarded.
     assert not any("ignored" in argv for argv in argvs)
+
+
+def run_returning(monkeypatch, stdout=b"", returncode=0):
+    """Run seaf-cli commands that produce the given output."""
+    calls = []
+
+    def fake_run(argv, **kwargs):
+        calls.append(argv)
+        return subprocess.CompletedProcess(argv, returncode, stdout=stdout)
+
+    monkeypatch.setattr(subprocess, "run", fake_run)
+    return calls
+
+
+def test_local_libraries_are_read_as_json(client, monkeypatch):
+    out = json.dumps([
+        {"name": "Docs", "id": "id-1", "path": "/dsc/seafile/Docs"},
+    ]).encode()
+    calls = run_returning(monkeypatch, out)
+
+    assert client.get_local_libraries() == {"id-1"}
+    assert "--json" in calls[-1]
+
+
+@pytest.mark.parametrize("name, path", [
+    # Display output is whitespace separated, so these used to be unparseable.
+    ("My Docs", "/dsc/seafile/My_Docs"),
+    ("Docs", "/mnt/my libraries/Docs"),
+    ("My Docs", "/mnt/my libraries/My Docs"),
+    ("Ünïcödé ☂", "/dsc/seafile/Ünïcödé_☂"),
+    ("tab\tname", "/dsc/seafile/tab_name"),
+])
+def test_local_libraries_survive_whitespace_in_names_and_paths(client, monkeypatch, name, path):
+    out = json.dumps([{"name": name, "id": "id-1", "path": path}]).encode()
+    run_returning(monkeypatch, out)
+
+    assert client.get_local_libraries() == {"id-1"}
+
+
+def test_no_local_libraries_gives_an_empty_set(client, monkeypatch):
+    run_returning(monkeypatch, b"[]")
+    assert client.get_local_libraries() == set()
+
+
+def test_local_libraries_reports_unreadable_output(client, monkeypatch):
+    run_returning(monkeypatch, b"not json at all")
+    with pytest.raises(DaemonError):
+        client.get_local_libraries()
+
+
+def test_local_libraries_reports_a_failed_command(client, monkeypatch):
+    run_returning(monkeypatch, b"", returncode=1)
+    with pytest.raises(DaemonError):
+        client.get_local_libraries()
+
+
+@pytest.mark.parametrize("payload", [
+    b'{"name": "Docs"}',            # an object where a list is expected
+    b'[{"name": "Docs"}]',          # an entry with no id
+    b'["Docs"]',                    # entries that are not objects
+    b'[null]',
+    b'null',
+    b'42',
+])
+def test_local_libraries_reports_unexpected_json_shapes(client, monkeypatch, payload):
+    """Valid JSON of the wrong shape must not escape as a raw TypeError."""
+    run_returning(monkeypatch, payload)
+    with pytest.raises(DaemonError):
+        client.get_local_libraries()
+
+
+def test_local_libraries_error_keeps_the_original_cause(client, monkeypatch):
+    run_returning(monkeypatch, b"not json at all")
+    with pytest.raises(DaemonError) as err:
+        client.get_local_libraries()
+    assert err.value.__cause__ is not None