Files
another-boids-in-rust/vendor/rustybuzz/scripts/gen-vowel-constraints.py

193 lines
7.1 KiB
Python
Executable File

#!/usr/bin/env python3
"""
Generator of the function to prohibit certain vowel sequences.
It creates ``preprocess_text_vowel_constraints``, which inserts dotted
circles into sequences prohibited by the USE script development spec.
Based on harfbuzz/src/gen-vowel-constraints.py
"""
import collections
import io
import os
import urllib.request
if not os.path.exists('Scripts.txt'):
urllib.request.urlretrieve('https://unicode.org/Public/14.0.0/ucd/Scripts.txt', 'Scripts.txt')
with io.open('Scripts.txt', encoding='utf-8') as f:
scripts_header = [f.readline() for i in range(2)]
scripts = {}
script_order = {}
for line in f:
j = line.find('#')
if j >= 0:
line = line[:j]
fields = [x.strip() for x in line.split(';')]
if len(fields) == 1:
continue
uu = fields[0].split('..')
start = int(uu[0], 16)
if len(uu) == 1:
end = start
else:
end = int(uu[1], 16)
script = fields[1]
for u in range(start, end + 1):
scripts[u] = script
if script not in script_order:
script_order[script] = start
class ConstraintSet(object):
"""A set of prohibited code point sequences.
Args:
constraint (List[int]): A prohibited code point sequence.
"""
def __init__(self, constraint):
# Either a list or a dictionary. As a list of code points, it
# represents a prohibited code point sequence. As a dictionary,
# it represents a set of prohibited sequences, where each item
# represents the set of prohibited sequences starting with the
# key (a code point) concatenated with any of the values
# (ConstraintSets).
self._c = constraint
def add(self, constraint):
"""Add a constraint to this set."""
if not constraint:
return
first = constraint[0]
rest = constraint[1:]
if isinstance(self._c, list):
if constraint == self._c[:len(constraint)]:
self._c = constraint
elif self._c != constraint[:len(self._c)]:
self._c = {self._c[0]: ConstraintSet(self._c[1:])}
if isinstance(self._c, dict):
if first in self._c:
self._c[first].add(rest)
else:
self._c[first] = ConstraintSet(rest)
def __str__(self, index=0, depth=4):
s = []
if isinstance(self._c, list):
if len(self._c) == 0:
assert index == 2, 'Cannot use `matched` for this constraint; the general case has not been implemented'
s.append('matched = true;\n')
elif len(self._c) == 1:
assert index == 1, 'Cannot use `matched` for this constraint; the general case has not been implemented'
s.append('matched = 0x{:04X} == buffer.cur({}).glyph_id;\n'.format(next(
iter(self._c)), index))
else:
s.append('if 0x{:04X} == buffer.cur({}).glyph_id &&\n'.format(self._c[0], index))
if index:
s.append('buffer.idx + {} < buffer.len &&\n'.format(index + 1))
for i, cp in enumerate(self._c[1:], start=1):
s.append('0x{:04X} == buffer.cur({}).glyph_id{}\n'.format(
cp, index + i, '' if i == len(self._c) - 1 else ' &&'))
s.append('{\n')
for i in range(index + 1):
s.append('buffer.next_glyph();\n')
s.append('output_dotted_circle(buffer);\n')
s.append('}\n')
else:
s.append('match buffer.cur({}).glyph_id {{\n'.format(index))
cases = collections.defaultdict(set)
for first, rest in sorted(self._c.items()):
cases[rest.__str__(index + 1, depth + 2)].add(first)
for ii, (body, labels) in enumerate(sorted(cases.items(), key=lambda b_ls: sorted(b_ls[1])[0])):
for i, cp in enumerate(sorted(labels)):
if i == len(labels) - 1:
s.append(' 0x{:04X} => {{ {}'.format(cp, '\n' if i % 4 == 3 else ''))
else:
s.append(' 0x{:04X} | {}'.format(cp, '\n' if i % 4 == 3 else ''))
s.append(body)
s.append('}')
if ii == len(cases.items()) - 1:
s.append('_ => {}')
s.append('}\n')
return ''.join(s)
constraints = {}
with io.open('ms-use/IndicShapingInvalidCluster.txt', encoding='utf-8') as f:
constraints_header = []
while True:
line = f.readline().strip()
if line == '#':
break
constraints_header.append(line)
for line in f:
j = line.find('#')
if j >= 0:
line = line[:j]
constraint = [int(cp, 16) for cp in line.split(';')[0].split()]
if not constraint:
continue
assert 2 <= len(constraint), 'Prohibited sequence is too short: {}'.format(constraint)
script = scripts[constraint[0]]
if script in constraints:
constraints[script].add(constraint)
else:
constraints[script] = ConstraintSet(constraint)
assert constraints, 'No constraints found'
print('// WARNING: this file was generated by scripts/gen-vowel-constraints.py')
print()
print('use super::buffer::hb_buffer_t;')
print('use super::ot_layout::*;')
print('use super::script;')
print('use crate::BufferFlags;')
print()
print('fn output_dotted_circle(buffer: &mut hb_buffer_t) {')
print(' buffer.output_glyph(0x25CC);')
print(' {')
print(' let out_idx = buffer.out_len - 1;')
print(' _hb_glyph_info_reset_continuation(&mut buffer.out_info_mut()[out_idx]);')
print(' }')
print('}')
print()
print('fn output_with_dotted_circle(buffer: &mut hb_buffer_t) {')
print(' output_dotted_circle(buffer);')
print(' buffer.next_glyph();')
print('}')
print()
print('pub fn preprocess_text_vowel_constraints(buffer: &mut hb_buffer_t) {')
print(' if buffer.flags.contains(BufferFlags::DO_NOT_INSERT_DOTTED_CIRCLE) {')
print(' return;')
print(' }')
print()
print(' // UGLY UGLY UGLY business of adding dotted-circle in the middle of')
print(' // vowel-sequences that look like another vowel. Data for each script')
print(' // collected from the USE script development spec.')
print(' //')
print(' // https://github.com/harfbuzz/harfbuzz/issues/1019')
print(' buffer.clear_output();')
print(' match buffer.script {')
for script, constraints in sorted(constraints.items(), key=lambda s_c: script_order[s_c[0]]):
print(' Some(script::{}) => {{'.format(script.upper()))
print(' buffer.idx = 0;')
print(' while buffer.idx + 1 < buffer.len {')
print(' #[allow(unused_mut)]')
print(' let mut matched = false;')
print(str(constraints), end='')
print(' buffer.next_glyph();')
print(' if matched { output_with_dotted_circle(buffer); }')
print(' }')
print(' }')
print()
print(' _ => {}')
print(' }')
print(' buffer.sync();')
print('}')
print()