harfbuzz/src/gen-ucd-table.py

166 lines
5.3 KiB
Python
Executable File

#!/usr/bin/env python3
import io, os.path, sys, re
import logging
logging.basicConfig(format='%(levelname)s: %(message)s', level=logging.INFO)
if len (sys.argv) not in (2, 3):
print("""usage: ./gen-ucd-table ucd.nounihan.grouped.xml [/path/to/hb-common.h]
Input file, as of Unicode 12:
* https://unicode.org/Public/UCD/latest/ucdxml/ucd.nounihan.grouped.zip""", file=sys.stderr)
sys.exit(1)
# https://github.com/harfbuzz/packtab
import packTab
import packTab.ucdxml
logging.info('Loading UCDXML...')
ucdxml = packTab.ucdxml.load_ucdxml(sys.argv[1])
ucd = packTab.ucdxml.ucdxml_get_repertoire(ucdxml)
hb_common_h = 'hb-common.h' if len (sys.argv) < 3 else sys.argv[2]
logging.info('Preparing data tables...')
gc = [u['gc'] for u in ucd]
ccc = [int(u['ccc']) for u in ucd]
bmg = [int(v, 16) - int(u) if v else 0 for u,v in enumerate(u['bmg'] for u in ucd)]
#gc_ccc_non0 = set((cat,klass) for cat,klass in zip(gc,ccc) if klass)
#gc_bmg_non0 = set((cat,mirr) for cat,mirr in zip(gc, bmg) if mirr)
sc = [u['sc'] for u in ucd]
dm = {i:tuple(int(v, 16) for v in u['dm'].split()) for i,u in enumerate(ucd)
if u['dm'] != '#' and u['dt'] == 'can' and not (0xAC00 <= i < 0xAC00+11172)}
ce = {i for i,u in enumerate(ucd) if u['Comp_Ex'] == 'Y'}
assert not any(v for v in dm.values() if len(v) not in (1,2))
dm1 = sorted(set(v for v in dm.values() if len(v) == 1))
assert all((v[0] >> 16) in (0,2) for v in dm1)
dm1_p0_array = ['0x%04Xu' % (v[0] & 0xFFFF) for v in dm1 if (v[0] >> 16) == 0]
dm1_p2_array = ['0x%04Xu' % (v[0] & 0xFFFF) for v in dm1 if (v[0] >> 16) == 2]
dm1_order = {v:i+1 for i,v in enumerate(dm1)}
dm2 = sorted((v+(i if i not in ce and not ccc[i] else 0,), v)
for i,v in dm.items() if len(v) == 2)
filt = lambda v: ((v[0] & 0xFFFFF800) == 0x0000 and
(v[1] & 0xFFFFFF80) == 0x0300 and
(v[2] & 0xFFF0C000) == 0x0000)
dm2_u32_array = [v for v in dm2 if filt(v[0])]
dm2_u64_array = [v for v in dm2 if not filt(v[0])]
assert dm2_u32_array + dm2_u64_array == dm2
dm2_u32_array = ["HB_CODEPOINT_ENCODE3_11_7_14 (0x%04Xu, 0x%04Xu, 0x%04Xu)" % v[0] for v in dm2_u32_array]
dm2_u64_array = ["HB_CODEPOINT_ENCODE3 (0x%04Xu, 0x%04Xu, 0x%04Xu)" % v[0] for v in dm2_u64_array]
l = 1 + len(dm1_p0_array) + len(dm1_p2_array)
dm2_order = {v[1]:i+l for i,v in enumerate(dm2)}
dm_order = {None: 0}
dm_order.update(dm1_order)
dm_order.update(dm2_order)
gc_order = dict()
for i,v in enumerate(('Cc', 'Cf', 'Cn', 'Co', 'Cs', 'Ll', 'Lm', 'Lo', 'Lt', 'Lu',
'Mc', 'Me', 'Mn', 'Nd', 'Nl', 'No', 'Pc', 'Pd', 'Pe', 'Pf',
'Pi', 'Po', 'Ps', 'Sc', 'Sk', 'Sm', 'So', 'Zl', 'Zp', 'Zs',)):
gc_order[i] = v
gc_order[v] = i
sc_order = dict()
sc_array = []
sc_re = re.compile(r"\b(HB_SCRIPT_[_A-Z]*).*HB_TAG [(]'(.)','(.)','(.)','(.)'[)]")
for line in open(hb_common_h):
m = sc_re.search (line)
if not m: continue
name = m.group(1)
tag = ''.join(m.group(i) for i in range(2, 6))
i = len(sc_array)
sc_order[tag] = i
sc_order[i] = tag
sc_array.append(name)
DEFAULT = 1
COMPACT = 3
SLOPPY = 5
logging.info('Generating output...')
print("/* == Start of generated table == */")
print("/*")
print(" * The following table is generated by running:")
print(" *")
print(" * ./gen-ucd-table.py ucd.nounihan.grouped.xml")
print(" *")
print(" * on file with this description:", ucdxml.description)
print(" */")
print()
print("#ifndef HB_UCD_TABLE_HH")
print("#define HB_UCD_TABLE_HH")
print()
print('#include "hb.hh"')
print()
code = packTab.Code('_hb_ucd')
sc_array, _ = code.addArray('hb_script_t', 'sc_map', sc_array)
dm1_p0_array, _ = code.addArray('uint16_t', 'dm1_p0_map', dm1_p0_array)
dm1_p2_array, _ = code.addArray('uint16_t', 'dm1_p2_map', dm1_p2_array)
dm2_u32_array, _ = code.addArray('uint32_t', 'dm2_u32_map', dm2_u32_array)
dm2_u64_array, _ = code.addArray('uint64_t', 'dm2_u64_map', dm2_u64_array)
code.print_c(linkage='static inline')
datasets = [
('gc', gc, 'Cn', gc_order),
('ccc', ccc, 0, None),
('bmg', bmg, 0, None),
('sc', sc, 'Zzzz', sc_order),
('dm', dm, None, dm_order),
]
for compression in (DEFAULT, COMPACT, SLOPPY):
logging.info(' Compression=%d:' % compression)
print()
if compression == DEFAULT:
print('#ifndef HB_OPTIMIZE_SIZE')
elif compression == COMPACT:
print('#elif !defined(HB_NO_UCD_UNASSIGNED)')
else:
print('#else')
print()
if compression == SLOPPY:
for i in range(len(gc)):
if (i % 128) and gc[i] == 'Cn':
gc[i] = gc[i - 1]
for i in range(len(gc) - 2, -1, -1):
if ((i + 1) % 128) and gc[i] == 'Cn':
gc[i] = gc[i + 1]
for i in range(len(sc)):
if (i % 128) and sc[i] == 'Zzzz':
sc[i] = sc[i - 1]
for i in range(len(sc) - 2, -1, -1):
if ((i + 1) % 128) and sc[i] == 'Zzzz':
sc[i] = sc[i + 1]
code = packTab.Code('_hb_ucd')
for name,data,default,mapping in datasets:
sol = packTab.pack_table(data, default, mapping=mapping, compression=compression)
logging.info(' Dataset=%-8s FullCost=%d' % (name, sol.fullCost))
sol.genCode(code, name)
code.print_c(linkage='static inline')
print()
print('#endif')
print()
print()
print("#endif /* HB_UCD_TABLE_HH */")
print()
print("/* == End of generated table == */")
logging.info('Done.')