Add an artifact-type suffix to standardized filenames (#1404)

Standardized artifact filenames gain a trailing type segment for the two
types whose file is ambiguous once downloaded: a talk's exported slides
and a poster generate the same Author_TitleInTitleCase_VenueYear name as
the paper itself. Talks now end in "_Talk", posters in "_Poster";
publications and grants stay unsuffixed (a bare paper PDF is the default
expectation, and that name is also the .bib download name).

Trailing rather than the existing mid-name suffix slot, so everything
before it stays byte-identical to the old scheme and the "-<timestamp>"
uniquifier still appends cleanly. The label comes from a class attribute
(FILENAME_TYPE_SUFFIX), not from data -- Poster has no type field and
Talk.talk_type is nullable and editor-editable, so deriving it would
rename files on a metadata-only edit.

Applied retroactively via restandardize_artifact_filenames, which is
already an idempotent every-container-start step. It ships here with
--dry-run so the scope is reviewable in the logs before any file moves.

The load-bearing subtlety is ordering: backfill_original_filenames runs
BEFORE that step, and its "already standardized?" guard is what stops it
recording a renamed file's current name as its "Originally uploaded as"
provenance. The moment the scheme changed, every already-renamed talk and
poster would have failed that guard and had its old standardized name
written in as a fake original -- so the guard now accepts both schemes.

Also extracts the uniquifier-tolerant name comparison that three commands
had each copy-pasted into one tested helper (matches_standardized_basename),
and teaches repair_diverged_artifact_filenames to look for orphans under
the pre-#1404 base too (that bug predates the scheme change).

Known and accepted: /media/talks/ and /media/posters/ 404s are answered by
Apache and never reach Django, so unlike publications (which serve_pdf can
redirect via the recorded original filename) external links to renamed
talk/poster PDFs cannot be rescued from this repo. Tracked separately.

Tests: 661 OK. New coverage for the trailing segment, per-class labels,
the shared matcher, the retroactive rename of a pre-#1404 talk (pdf + raw
+ thumbnail) and its idempotency, publications not churning, and the
backfill guard for both a legacy name and its uniquified variant.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Author: jonfroehlich

docker-entrypoint.sh

@@ -164,9 +164,15 @@ echo "******************************************"
 python manage.py setup_admin_groups
 
 echo "****************** STEP 4.10b/5: docker-entrypoint.sh ************************"
-echo "4.10b Running 'python manage.py restandardize_artifact_filenames' to rename legacy talk/poster/pub files to the standardized scheme (#1401)"
-echo "******************************************"
-python manage.py restandardize_artifact_filenames
+echo "4.10b Running 'python manage.py restandardize_artifact_filenames' to rename legacy talk/poster/pub files to the standardized scheme (#1401/#1404)"
+echo "******************************************"
+# TEMPORARY (#1404): the scheme just gained a trailing artifact-type segment
+# ("..._CHI2024_Talk"), so this step would re-rename EVERY already-standardized
+# talk and poster in one unattended pass. --dry-run logs what WOULD be renamed,
+# touching nothing on disk or in the DB, so the scope can be reviewed on the
+# test server (and then prod) first. REMOVE --dry-run and redeploy to perform
+# the rename; after that this step is idempotent again and stays in place.
+python manage.py restandardize_artifact_filenames --dry-run
 
 echo "****************** STEP 4.10c/5: docker-entrypoint.sh ************************"
 echo "4.10c Running 'python manage.py repair_diverged_artifact_filenames' to recover artifacts whose files were renamed on disk but not in the DB (#1390 dotted-name bug)"

makeabilitylab/settings.py

@@ -86,8 +86,8 @@
     SECURE_PROXY_SSL_HEADER = ('HTTP_X_FORWARDED_PROTO', 'https')
 
 # Makeability Lab Global Variables, including Makeability Lab version
-ML_WEBSITE_VERSION = "2.28.1" # Keep this updated with each release and also change the short description below
-ML_WEBSITE_VERSION_DESCRIPTION = "Adds two quick links to the admin header so a pair of easy-to-forget internal pages stop being invisible: Project People (/view-project-people/, the project-team roster, which nothing in the main site nav links to) and Activity Log (the site-wide admin action log added in #1413). The Activity Log link is wrapped in {% if user.is_superuser %} so it matches the gate on LogEntryAdmin itself -- editors and contributors never see a link to a page they cannot open, and the log's existence is not advertised to them. The roster link is labeled 'Project People' to match that page's own heading rather than inventing a second name for the same page, and both new-tab links now carry a visually-hidden '(opens in a new tab)' cue built on Django admin's own screen-reader helper class, so no new CSS ships. Also fixes a NoReverseMatch this change would otherwise have shipped: website/urls.py sets app_name = 'website', so the roster reverses as 'website:view_project_people', and the bare name raised inside the userlinks block of base_site.html -- the parent template of EVERY admin page -- which would have 500'd the whole admin for any logged-in staff user rather than merely breaking one link (and, on -test where DEBUG is True, shown a public traceback). It slipped past manual testing because hitting /view-project-people/ exercises the URL path rather than the reverse name, and an unauthenticated admin request 302s to login without ever rendering the template. A new regression test renders the admin index as both a superuser and a staff non-superuser, pinning both reverses, the superuser gate, and the new-tab cue, so a future URL-name rename fails a test instead of taking down the admin."
+ML_WEBSITE_VERSION = "2.29.0" # Keep this updated with each release and also change the short description below
+ML_WEBSITE_VERSION_DESCRIPTION = "Standardized filenames for talks and posters now end in '_Talk' and '_Poster' (#1404), so a downloaded slide deck or poster is no longer indistinguishable from the paper PDF of the same work. Publication filenames are unchanged. Existing talk/poster files are renamed once, on deploy."
 DATE_MAKEABILITYLAB_FORMED = datetime.date(2012, 1, 1)  # Date Makeability Lab was formed
 MAX_BANNERS = 7 # Maximum number of banners on a page
 

website/management/commands/backfill_original_filenames.py

@@ -4,6 +4,7 @@
 from django.core.management.base import BaseCommand
 
 from website.models import Artifact, Talk, Poster, Publication
+from website.utils import fileutils as ml_fileutils
 
 # This retrieves a Python logging instance (or creates it)
 _logger = logging.getLogger(__name__)
@@ -122,19 +123,26 @@ def _backfill_row(self, model, artifact, file_attr, original_attr, dry_run):
 
         current_basename = os.path.basename(file_field.name)
         current_no_ext = os.path.splitext(current_basename)[0]
-        standardized_no_ext = Artifact.generate_filename(artifact)
-
-        # Treat the file as already-standardized when its name equals the
-        # standardized scheme OR is a uniquified variant of it. When a
-        # standardized name collides on disk, ensure_filename_is_unique()
-        # (fileutils.py) appends "-<timestamp>" — e.g.
-        # "Lee_Talk_CHI2021-1782399772.42.pdf" — so the on-disk name still
-        # STARTS WITH the standardized base. Matching only on exact equality
-        # would misread those as never-renamed and record the standardized+
-        # suffix name as the "original" — a false positive.
-        already_standardized = (
-            current_no_ext == standardized_no_ext
-            or current_no_ext.startswith(standardized_no_ext + "-")
+
+        # Treat the file as already-standardized when its name matches the
+        # standardized scheme. matches_standardized_basename also accepts the
+        # "-<timestamp>" that ensure_filename_is_unique appends on a disk
+        # collision (e.g. "Lee_Talk_CHI2021-1782399772.42.pdf"), so a renamed-
+        # then-uniquified file isn't misread as an original upload.
+        #
+        # BOTH schemes count (#1404). This command runs at container start
+        # BEFORE restandardize_artifact_filenames, so on the deploy that
+        # introduces the artifact-type segment ("..._Talk") every already-renamed
+        # talk and poster still carries its pre-#1404 name. Checking only the
+        # current scheme would read the whole corpus as never-renamed and record
+        # those old standardized names as originals — false provenance, shown to
+        # editors as "Originally uploaded as".
+        already_standardized = any(
+            ml_fileutils.matches_standardized_basename(
+                current_no_ext,
+                Artifact.generate_filename(
+                    artifact, include_type_suffix=include_type_suffix))
+            for include_type_suffix in (True, False)
         )
         if already_standardized:
             # Already renamed — the original upload name is gone.

website/management/commands/repair_diverged_artifact_filenames.py

@@ -4,6 +4,7 @@
 from django.core.management.base import BaseCommand
 
 from website.models import Artifact, Talk, Poster, Publication
+from website.utils import fileutils as ml_fileutils
 
 # This retrieves a Python logging instance (or creates it)
 _logger = logging.getLogger(__name__)
@@ -150,9 +151,15 @@ def _repair_field(self, artifact, field_name, dry_run):
         # base, possibly with a "-<timestamp>" uniqueness suffix from colliding
         # with the sibling files. Match those, then confirm by content type so we
         # never mis-pair a pdf with a pptx/thumbnail (they share the base name).
+        # The pre-#1404 base (no trailing "_Talk"/"_Poster") is searched too: the
+        # bug predates that scheme change, so an orphan it left behind on disk
+        # carries the old base even though the repair target uses the new one.
+        legacy_base = get_valid_filename(
+            Artifact.generate_filename(artifact, include_type_suffix=False))
         candidates = []
         for entry in os.listdir(directory):
-            if entry == valid_base or entry.startswith(valid_base + "-"):
+            if any(ml_fileutils.matches_standardized_basename(entry, base)
+                   for base in {valid_base, legacy_base}):
                 full = os.path.join(directory, entry)
                 if os.path.isfile(full):
                     candidates.append(entry)

website/management/commands/restandardize_artifact_filenames.py

@@ -4,24 +4,28 @@
 from django.core.management.base import BaseCommand
 
 from website.models import Artifact, Talk, Poster, Publication
+from website.utils import fileutils as ml_fileutils
 
 # This retrieves a Python logging instance (or creates it)
 _logger = logging.getLogger(__name__)
 
 
 class Command(BaseCommand):
     help = (
-        "Re-standardizes legacy talk/poster/publication filenames that were "
-        "never renamed to the Author_TitleInTitleCase_VenueYear scheme (issue "
-        "#1401). Production has many such rows (bulk-imported, so they never "
-        "went through an authored Artifact.save()). This reuses the existing, "
-        "now-correct rename path: when Artifact.do_filenames_need_updating() is "
-        "True it calls artifact.save(), which renames the pdf_file, raw_file, "
-        "and thumbnail on disk AND in the DB together. The original upload name "
-        "is preserved (it was captured into original_*_filename by "
-        "backfill_original_filenames / #1391 before this runs). Idempotent: "
-        "once a row is standardized the check returns False, so re-runs do "
-        "nothing. Safe to run on every container start."
+        "Re-standardizes talk/poster/publication filenames that don't match the "
+        "current scheme. Two populations: files that were never renamed at all "
+        "(issue #1401 — bulk-imported rows that never went through an authored "
+        "Artifact.save()), and files standardized under a superseded scheme "
+        "(issue #1404 added the trailing artifact-type segment, e.g. "
+        "'..._CHI2024_Talk'), which this migrates in one pass. It reuses the "
+        "existing, now-correct rename path: when Artifact.do_filenames_need_"
+        "updating() is True it calls artifact.save(), which renames the "
+        "pdf_file, raw_file, and thumbnail on disk AND in the DB together. The "
+        "original upload name is preserved (it was captured into "
+        "original_*_filename by backfill_original_filenames / #1391 before this "
+        "runs). Idempotent: once a row matches the current scheme the check "
+        "returns False, so re-runs do nothing. Safe to run on every container "
+        "start."
     )
 
     # The concrete artifact models this covers. Posters are already
@@ -156,9 +160,12 @@ def _needs_restandardizing(artifact):
         standardized name collided on disk and got a ``-<timestamp>`` suffix
         (``ensure_filename_is_unique``) reads as "needs updating" forever and
         would be re-renamed on every run — churning duplicate-name artifacts'
-        filenames on every deploy. Here a name that equals the standardized
-        base OR is a ``-<suffix>`` variant of it counts as already standardized,
-        which keeps the command idempotent.
+        filenames on every deploy. ``matches_standardized_basename`` accepts
+        those variants, which keeps the command idempotent.
+
+        Only the CURRENT scheme counts as standardized: a file named under the
+        pre-#1404 scheme (no trailing "_Talk"/"_Poster") is deliberately flagged,
+        which is what migrates the existing corpus in a single pass.
         """
         standardized = Artifact.generate_filename(artifact)
         for file_attr in ("pdf_file", "raw_file"):
@@ -167,10 +174,7 @@ def _needs_restandardizing(artifact):
                 continue
             current_no_ext = os.path.splitext(
                 os.path.basename(file_field.name))[0]
-            is_standardized = (
-                current_no_ext == standardized
-                or current_no_ext.startswith(standardized + "-")
-            )
-            if not is_standardized:
+            if not ml_fileutils.matches_standardized_basename(
+                    current_no_ext, standardized):
                 return True
         return False

website/models/artifact.py

@@ -23,6 +23,14 @@ class Artifact(models.Model):
       2. also hook up an authors_changed signal in signals.py. For example, add the line "@receiver(m2m_changed, sender=Grant.authors.through)"
          to signals.py's def authors_changed(sender, instance, action, reverse, **kwargs):
     """
+    # The artifact-type segment appended to standardized filenames (#1404), so a
+    # downloaded file says what kind of artifact it is. Set on the subclasses
+    # whose files are otherwise ambiguous — a talk's exported slides and a poster
+    # generate the same Author_Title_VenueYear name as the paper itself. Left
+    # None here (and on Publication/Grant): a bare paper PDF is the default
+    # expectation, and that same name is also the .bib download name.
+    FILENAME_TYPE_SUFFIX = None
+
     title = models.CharField(max_length=255, blank=True, null=True)
     authors = SortedManyToManyField('Person', blank=True)
     date = models.DateField(null=True)
@@ -473,32 +481,44 @@ def do_filenames_need_updating(artifact):
         return False
 
     @staticmethod
-    def generate_filename(artifact, file_extension=None, max_pub_title_length = -1):
+    def generate_filename(artifact, file_extension=None, max_pub_title_length = -1,
+                          include_type_suffix=True):
         """
         Generates a filename for the given artifact.
 
-        This method generates a filename based on the artifact's first author's last name, title, forum name, and date.
+        This method generates a filename based on the artifact's first author's last name, title, forum name, and date,
+        plus the artifact-type segment of its class (``FILENAME_TYPE_SUFFIX``, e.g. "_Talk" — see #1404).
         If a file extension is provided, it is appended to the filename. Otherwise, a filename without extension is returned.
 
         Parameters:
         artifact (Artifact): The artifact for which the filename is to be generated.
         file_extension (str, optional): The file extension to be appended to the filename. Defaults to None.
+        include_type_suffix (bool, optional): Pass False for the pre-#1404 name (no type segment). Only the
+            filename management commands need this, to recognize a file that was standardized under the old
+            scheme; everything else wants the current scheme. Defaults to True.
 
         Returns:
         str: The generated filename.
 
         Example:
-        >>> artifact = Artifact(first_author_last_name="Froehlich", title="Research Artifact Title", forum_name="CHI", date="2023-12-16")
-        >>> generate_filename(artifact, file_extension=".pdf")
+        >>> talk = Talk(title="Research Artifact Title", forum_name="CHI", date="2023-12-16")  # first author: Froehlich
+        >>> generate_filename(talk, file_extension=".pdf")
+        'Froehlich_ResearchArtifactTitle_CHI2023_Talk.pdf'
+        >>> generate_filename(talk, file_extension=".pdf", include_type_suffix=False)
         'Froehlich_ResearchArtifactTitle_CHI2023.pdf'
         """
-        
+
+        # Read off the class, not the instance, so the type segment is a property of the model
+        # (Poster has no type field, and Talk.talk_type is nullable and editor-editable — deriving
+        # the segment from data would rename files on a metadata-only edit).
+        type_suffix = type(artifact).FILENAME_TYPE_SUFFIX if include_type_suffix else None
+
         # An empty string or a string with only whitespace characters is considered False in a boolean context.
         if not file_extension or not file_extension.strip():
             return ml_fileutils.get_filename_without_ext_for_artifact(
-                    artifact.get_first_author_last_name(), artifact.title, 
-                    artifact.forum_name, artifact.date)
+                    artifact.get_first_author_last_name(), artifact.title,
+                    artifact.forum_name, artifact.date, type_suffix=type_suffix)
         else:
             return ml_fileutils.get_filename_for_artifact(
-                    artifact.get_first_author_last_name(), artifact.title, 
-                    artifact.forum_name, artifact.date, file_extension)
+                    artifact.get_first_author_last_name(), artifact.title,
+                    artifact.forum_name, artifact.date, file_extension, type_suffix=type_suffix)

website/models/poster.py

@@ -6,6 +6,10 @@ class Poster(Artifact):
     UPLOAD_DIR = 'posters/'
     THUMBNAIL_DIR = os.path.join(UPLOAD_DIR, 'images/')
 
+    # Standardized filenames end in "_Poster" (#1404) so a downloaded poster
+    # isn't indistinguishable from the paper's PDF of the same work.
+    FILENAME_TYPE_SUFFIX = 'Poster'
+
     external_slides_url = models.URLField(blank=True, null=True)
     external_slides_url.help_text = (
         "Optional link to the source design (e.g., Figma, Canva, Illustrator "

website/models/talk.py

@@ -23,6 +23,10 @@ class Talk(Artifact):
     UPLOAD_DIR = 'talks/'
     THUMBNAIL_DIR = os.path.join(UPLOAD_DIR, 'images/')
 
+    # Standardized filenames end in "_Talk" (#1404) so a downloaded slide deck
+    # isn't indistinguishable from the paper's PDF of the same work.
+    FILENAME_TYPE_SUFFIX = 'Talk'
+
     external_slides_url = models.URLField(blank=True, null=True)
     external_slides_url.help_text = (
         "Optional link to the source slide deck (e.g., Figma, Google Slides, "