urwid / urwid / 37016855127

90%

push

github

Linear-time text layout; fix mixed markup runs (#1386)

Laying out a paragraph without spaces took time quadratic in its length,
and tag markup mixing str and bytes produced wrong attribute runs.

- calc_string_text_pos, calc_text_pos, is_wide_char, move_next_char and
  the UTF-8 grapheme decoder segment a window that doubles until the
  result is found, instead of the whole rest of the text.
- StandardTextLayout checks a line by the part that fits, and searches
  for each newline and tab once per paragraph, not per display line.
- move_prev_char stops at start_offs, and joining a line with the one
  before it must move the break past the removed space: a space plus a
  combining mark or emoji modifier no longer raises ValueError or makes
  the layout loop forever.
- decompose_tagmarkup converts elements of the other text type with
  errors="replace", as apply_target_encoding does, and recounts the
  attribute runs in the converted text.

| what | avg time before | avg time after | percent change |
| --- | --- | --- | --- |
| urwid.StandardTextLayout.layout, "x" * 80000, width 80, space | 5421.63 ms | 64.67 ms | -98.8% |
| urwid.StandardTextLayout.layout, U+00E9 * 40000, width 80, space | 4108.15 ms | 944.65 ms | -77.0% |
| urwid.StandardTextLayout.layout, b"x" * 40000 UTF-8, width 80, space | 1511.53 ms | 64.97 ms | -95.7% |
| urwid.Text.render, "x" * 40000, (80,) | 1431.40 ms | 34.82 ms | -97.6% |
| urwid.StandardTextLayout.layout, 80 lines of 60 chars, width 80 | 39.0 us | 40.7 us | +4.4% |

4603 of 5538 branches covered (83.12%)

Branch coverage included in aggregate %.

85 of 90 new or added lines in 3 files covered. (94.44%)

11558 of 13190 relevant lines covered (87.63%)

9.79 hits per line