Skip to content

INW000 unsupported-encoding

What it does

Reports a file in a layer or a context whose PEP 263 declaration (# coding: ... on line 1, or on line 2 after a blank or comment line) names a codec Inwards doesn't read. UTF-8 and its variants, ASCII, Latin-1, ISO-8859-* and cp125x are read as usual, because none of their bytes can turn into a line break or a quote. Any other codec, such as unicode_escape, utf-7, utf-16 or shift_jis, gets one INW000 on line 1, and the file's imports are not checked.

Why is this bad

Under unicode_escape, the text #\u000aimport shop.infrastructure.db is a comment to any tool that reads the file as UTF-8, and an import to CPython, which decodes \u000a to a line break first. An agent looking for a way past the layer check, or copying a trick from its training data, can hide any import that way. Inwards reads files as UTF-8 and ships no codec tables, so it reports the file rather than pass it unchecked (ADR-014).

Example

With the layers from INW001:

shop/domain/order.py
# -*- coding: unicode_escape -*-
#\u000aimport shop.infrastructure.db

inwards check

shop/domain/order.py:1:1: INW000 The file declares encoding "unicode-escape", which can hide imports from Inwards, so its imports were not checked.
  fix: Save the file as UTF-8 and remove the coding declaration.
    1. Re-encode the file as UTF-8 without changing its code.
    2. Delete the `# coding: ...` line, or change it to `# coding: utf-8`.
  docs: https://sircypkowskyy.github.io/inwards/rules/INW000/

Fixed: the file is UTF-8, and a comment is a comment again.

shop/domain/order.py
# coding: utf-8
"""Orders."""

inwards check

How to fix

  1. Re-encode the file as UTF-8 without changing its code (iconv -f SHIFT_JIS -t UTF-8, or your editor's "save with encoding").
  2. Delete the # coding: ... line, or change it to # coding: utf-8.
  3. Run the check again: the file's imports are checked now, so an import that the old encoding hid shows up as INW001 or another rule.

Fix safety

No automatic fix (autofix: false). Re-encoding has to keep every character the same, which a person should check for files with non-ASCII text.

Configuration

Always an error. [tool.inwards.rules] can't list INW000 in ignore or severity (a config error, exit 2), and select doesn't turn it off: a file whose declared encoding can hide imports isn't checked at all, so turning INW000 off would let it pass unchecked (ADR-027). It can't be suppressed inline either, and a file with INW000 has no suppressions read, because under its declared encoding a "comment" may be code (ADR-028).

Known limitations

  • Legitimate files in codecs such as Shift_JIS or EUC-JP are reported too. That is deliberate: a Shift_JIS lead byte can swallow a backslash.
  • Only files whose imports are checked get INW000: files in a layer, and every file once contexts are declared. Any other file with such a declaration isn't reported, though INW007 and INW009 still check it.
  • Bytes passed to exec or compile with an unreadable declared encoding are reported by INW011, not INW000.

References