[appleweiping]

Summary

• remove the unused prettyprint package from the base installation
• constrain PrettyPrintTree to >=1.36.4,<2, whose API matches the existing PortTree.print() call
• install that renderer in the lightweight unit-test job, which intentionally does not install the package's full dependency set
• add a regression test that verifies PortTree.print(savetofile=True) writes the expected hierarchy

Why

A clean Python 3.11 installation of current main resolves both prettyprint 0.1.5 and prettyprinttree 2.0.2. On the case-insensitive Windows filesystem used for reproduction, their prettyprint/ and PrettyPrint/ package trees collide. The imports in port_utils.py then fall through to the no-op shim, so PortTree.print() silently produces no file.

Removing only the unused package is not sufficient: PrettyPrintTree 2.0.2 changed its constructor API and raises TypeError for the existing default_orientation option. Version 1.36.4 has the compatible API and already ships a universal py3-none-any wheel.

The original report concerned Apple Silicon. I have not reproduced on arm64 macOS, so this PR limits its claim to the dependency conflict and API mismatch verified on Windows/Python 3.11.

Tests

• python -m pytest -q tests/test_port_tree.py — 1 passed
• python -m pytest -q tests/test_port_tree.py tests/test_import_paths.py tests/test_repo_layout.py — 9 passed
• python -m pytest -q tests --ignore=tests/test_cells_layout.py — 26 passed, 16 subtests passed
• .github/workflows/tests.yml parsed successfully with PyYAML
• regression check with the dependencies selected by current setup.py (prettyprint 0.1.5 + PrettyPrintTree 2.0.2) — new test fails because the output file is not created

The passing runs report existing Pydantic v2 deprecation warnings; this change does not add or address them.

Review status

This is intentionally a draft while awaiting maintainer direction in #78. It preserves the current PortTree API and avoids broader renderer or packaging refactors.

Closes #78

Assisted-by: OpenAI Codex

[appleweiping]

Closing this draft in favour of #123, which supersedes its dependency fix.

This draft pinned prettyprinttree>=1.36.4,<2. That alone is not sufficient: 1.36.x does not constrain its own cmd2 dependency, and cmd2 4.x moved cmd2.ansi, which makes PrettyPrint unimportable — port_utils then silently falls back to its no-op shim, so PortTree.print() produces no output rather than raising. #123 adds the cmd2<3 bound as well, so the printer actually runs.

This draft also carries a regression test (tests/test_port_tree.py) that builds a component, calls PortTree.print(savetofile=True, ...), and asserts the hierarchy reaches the file, plus a line adding prettyprinttree to the test job so the fast gdstk workflow would cover it. I have not kept that in #123 because I cannot run gLayout locally to confirm the test passes in your CI environment, and I would rather not add a test to a PR on the strength of an untested guess. If you would like it there, say so on #123 and I will add it — or take it from this branch directly.