[escaped]

Why

Fix #23. get_orientation crashed with KeyError: 'Orientation' for any
file whose exif data has no orientation tag. ExifTool omits the tag when an
image is stored unrotated (common for screenshots, PNGs and processed
images), so ExifField(denormalized_fields={'orientation': get_orientation})
broke on perfectly valid files.

Walkthrough

• Default the orientation to 1 (normal) when the tag is missing, instead of
  assuming it is present.
• Add a regression test covering a missing tag for both landscape and
  portrait dimensions.

Changes

File(s)	Change Summary
exiffield/getters.py	read Orientation with a normal-orientation default
tests/test_getters.py	regression test without an Orientation tag
CHANGELOG.md	note the fix under [Unreleased]

Testing

Red first: the new test raised KeyError: 'Orientation' before the fix
(2 failed). After:

$ uv run pytest -q
71 passed, 1 xfailed in 5.80s
$ uv run ruff format --check . && uv run ruff check . && uv run mypy
17 files already formatted / All checks passed! / no issues found

CI on 3bb0c90: run 34694220972 — 14/14 jobs success.

PR breakdown

Stacked on the modernization branch test/e2e (PR #26). Merge in order.

#	PR	What it does	Status
1	This PR	#23: missing orientation tag crash	Open
2	#28	#17: timezone-aware get_datetaken	Open
3	#29	#6: document non-editable fields	Open
4	#30	#1: document and test deferred extraction	Open

[coveralls]

Coverage Report for CI Build 34694220972

Warning

No base build found for commit 0a11a59 on test/e2e.
Coverage changes can't be calculated without a base build.
If a base build is processing, this comment will update automatically when it completes.

Coverage: 98.276%

Details

• Patch coverage: 1 of 1 lines across 1 file are fully covered (100%).

Uncovered Changes

No uncovered changes found.

Coverage Regressions

Requires a base build to compare against. How to fix this →

---

Coverage Stats

Relevant Lines:	174
Covered Lines:	171
Line Coverage:	98.28%
Coverage Strength:	10.81 hits per line

---

💛 - Coveralls

[escaped]

Two-axis review record

Method: an independent sub-agent reviewed the four stacked diffs (this PR
through #30) on two axes — Standards (repo conventions, correctness,
test quality) and Spec (does each PR actually fix its issue as
reported) — and verified every finding by running the code. The findings
were fixed in follow-up commits; final state: 79 passed, 0 xfailed, ruff
format/check and mypy clean, all four branches CI-green (14/14 jobs each).

Findings addressed

PR	Finding	Disposition
#28	A present but malformed offset was silently ignored, returning a naive datetime even though the file claims to carry timezone info	Fixed in 48cb745: raises ExifError (a valid GPSDateTime is still preferred when available)
#29	The editable-copy workaround read self.exif before super().save(), but extraction runs in the field's pre_save — so it never seeded on the first save, and a cleared field was re-seeded	Fixed in bd00ac2: seed after the first save, guarded by created, preserving user edits including clearing
#30	README claimed commit=True fills denormalized_fields but the test model had none	Fixed in 7972298: UnsyncedImage gains a denormalized camera field, asserted after the deferred save
#30	Cleanup unlinked only the base file name, leaking storage-suffixed files across re-runs	Fixed in 7972298: delete the actual stored name
—	Docs PRs carry no CHANGELOG entry	Accepted: docs-only, consistent with repo history

Verified correct

• Every new test was observed failing before its fix (red runs are recorded
  in each PR's Testing section) — no vacuous assertions.
• get_datetaken preference order and fallbacks: offset-aware original →
  GPS (UTC, with or without Z) → naive original → ExifError; malformed
  offset falls back to GPS when present.
• get_orientation default of 1 (normal) is more correct than the
  issue's suggested 0: both bypass the 90° rotation branch, but 1 is the
  valid EXIF value.