From cc2f83dac08eb0add166a5635f8c5f2af4bd8c4e Mon Sep 17 00:00:00 2001
From: mvdbeek <m.vandenbeek@gmail.com>
Date: Fri, 24 May 2024 18:26:18 +0200
Subject: [PATCH] Raise appropriate exception if accessing deleted input file

---
 lib/galaxy/webapps/galaxy/api/job_files.py | 21 +++++++++++++++++----
 test/integration/test_job_files.py         | 19 ++++++++++++++++++-
 2 files changed, 35 insertions(+), 5 deletions(-)

diff --git a/lib/galaxy/webapps/galaxy/api/job_files.py b/lib/galaxy/webapps/galaxy/api/job_files.py
index df2edf2415cb..fa30c3433e48 100644
--- a/lib/galaxy/webapps/galaxy/api/job_files.py
+++ b/lib/galaxy/webapps/galaxy/api/job_files.py
@@ -11,6 +11,7 @@
     exceptions,
     util,
 )
+from galaxy.managers.context import ProvidesAppContext
 from galaxy.model import Job
 from galaxy.web import (
     expose_api_anonymous_and_sessionless,
@@ -34,7 +35,7 @@ class JobFilesAPIController(BaseGalaxyAPIController):
     """
 
     @expose_api_raw_anonymous_and_sessionless
-    def index(self, trans, job_id, **kwargs):
+    def index(self, trans: ProvidesAppContext, job_id, **kwargs):
         """
         GET /api/jobs/{job_id}/files
 
@@ -56,9 +57,21 @@ def index(self, trans, job_id, **kwargs):
         :rtype:     binary
         :returns:   contents of file
         """
-        self.__authorize_job_access(trans, job_id, **kwargs)
-        path = kwargs.get("path", None)
-        return open(path, "rb")
+        job = self.__authorize_job_access(trans, job_id, **kwargs)
+        path = kwargs["path"]
+        try:
+            return open(path, "rb")
+        except FileNotFoundError:
+            # We know that the job is not terminal, but users (or admin scripts) can purge input datasets.
+            # Here we discriminate that case from truly unexpected bugs.
+            # Not failing the job here, this is or should be handled by pulsar.
+            match = re.match(r"(galaxy_)?dataset_(.*)\.dat", os.path.basename(path))
+            if match:
+                # This looks like a galaxy dataset, check if any job input has been deleted.
+                if any(jtid.dataset.dataset.purged for jtid in job.input_datasets):
+                    raise exceptions.ItemDeletionException("Input dataset(s) for job have been purged.")
+            else:
+                raise
 
     @expose_api_anonymous_and_sessionless
     def create(self, trans, job_id, payload, **kwargs):
diff --git a/test/integration/test_job_files.py b/test/integration/test_job_files.py
index 36babdcebf6e..fce3bf4bd6bd 100644
--- a/test/integration/test_job_files.py
+++ b/test/integration/test_job_files.py
@@ -43,6 +43,7 @@
 
 class TestJobFilesIntegration(integration_util.IntegrationTestCase):
     initialized = False
+    dataset_populator: DatasetPopulator
 
     @classmethod
     def handle_galaxy_config_kwds(cls, config):
@@ -60,7 +61,7 @@ def setUp(self):
             sa_session = self.sa_session
             stmt = select(model.HistoryDatasetAssociation)
             assert len(sa_session.scalars(stmt).all()) == 0
-            self.dataset_populator.new_dataset(history_id, content=TEST_INPUT_TEXT, wait=True)
+            self.input_hda_dict = self.dataset_populator.new_dataset(history_id, content=TEST_INPUT_TEXT, wait=True)
             assert len(sa_session.scalars(stmt).all()) == 1
             self.input_hda = sa_session.scalars(stmt).all()[0]
             TestJobFilesIntegration.initialized = True
@@ -86,6 +87,22 @@ def test_read_by_state(self):
         response = requests.get(get_url, params=data)
         _assert_insufficient_permissions(response)
 
+    def test_read_fails_if_input_file_purged(self):
+        job, _, _ = self.create_static_job_with_state("running")
+        job_id, job_key = self._api_job_keys(job)
+        input_file_path = self.input_hda.get_file_name()
+        data = {"path": input_file_path, "job_key": job_key}
+        get_url = self._api_url(f"jobs/{job_id}/files", use_key=True)
+        head_response = requests.head(get_url, params=data)
+        api_asserts.assert_status_code_is_ok(head_response)
+        delete_response = self.dataset_populator.delete_dataset(
+            self.input_hda_dict["history_id"], content_id=self.input_hda_dict["id"], purge=True, wait_for_purge=True
+        )
+        assert delete_response.status_code == 200
+        head_response = requests.get(get_url, params=data)
+        assert head_response.status_code == 400
+        assert head_response.json()["err_msg"] == "Input dataset(s) for job have been purged."
+
     def test_write_by_state(self):
         job, output_hda, working_directory = self.create_static_job_with_state("running")
         job_id, job_key = self._api_job_keys(job)