Apparently string slices are not as smart as I was hoping for.
Slicing a looong (say, 1MB) string and holding onto it is not a
good idea if done thousands of times. So, do fewer slicings and
decompile subtables immediately instead of holding onto data.
This makes me want to rethink the kind of data structures we use
for lazy processing.
Fixes https://github.com/behdad/fonttools/issues/317
self.data is usually set by decompileHeader as "the data after the header"; while here
it also included `headerdata`. This produced corrupt data if the compile method was called
again, as the header data is added again to self.data.
Moreover, in none of the subtable classes' `compile` methods, the re-compiled data is
stored in self.data, so we shall not do that for format 14 either.
Fixes#389
Ie, now we get:
...
assert 0 <= value < 0x10000, value
AssertionError: (None, 'LookupListIndex', 'SubstLookupRecord[0]', 'ChainContextSubst[1]', 'Lookup[3]', 'LookupList')
The common stacktrace like this:
File "fonttools/Lib/fontTools/ttLib/__init__.py", line 202, in save
self._writeTable(tag, writer, done)
File "fonttools/Lib/fontTools/ttLib/__init__.py", line 631, in _writeTable
tabledata = self.getTableData(tag)
File "fonttools/Lib/fontTools/ttLib/__init__.py", line 644, in getTableData
return self.tables[tag].compile(self)
File "fonttools/Lib/fontTools/ttLib/tables/otBase.py", line 86, in compile
self.table.compile(writer, font)
File "fonttools/Lib/fontTools/ttLib/tables/otBase.py", line 681, in compile
conv.write(writer, font, table, value)
File "fonttools/Lib/fontTools/ttLib/tables/otConverters.py", line 354, in write
value.compile(subWriter, font)
File "fonttools/Lib/fontTools/ttLib/tables/otBase.py", line 661, in compile
conv.write(writer, font, table, value, i)
File "fonttools/Lib/fontTools/ttLib/tables/otConverters.py", line 354, in write
value.compile(subWriter, font)
File "fonttools/Lib/fontTools/ttLib/tables/otBase.py", line 661, in compile
conv.write(writer, font, table, value, i)
File "fonttools/Lib/fontTools/ttLib/tables/otConverters.py", line 354, in write
value.compile(subWriter, font)
File "fonttools/Lib/fontTools/ttLib/tables/otBase.py", line 661, in compile
conv.write(writer, font, table, value, i)
File "fonttools/Lib/fontTools/ttLib/tables/otConverters.py", line 277, in write
value.compile(writer, font)
File "fonttools/Lib/fontTools/ttLib/tables/otBase.py", line 681, in compile
conv.write(writer, font, table, value)
File "fonttools/Lib/fontTools/ttLib/tables/otConverters.py", line 175, in write
writer.writeUShort(value)
File "fonttools/Lib/fontTools/ttLib/tables/otBase.py", line 457, in writeUShort
assert 0 <= value < 0x10000, value
AssertionError: None
now has this as the last line:
AssertionError: (None, 'LookupListIndex', 'SubstLookupRecord', 'ChainContextSubst', 'Lookup', 'LookupList')
which means a value of None was tried for writing a LookupListIndex from a ChainContextSubset...
It's a hack, but a very useful one.
This reverts commit d8803873f6d4efbcb227867fe0ac6bca370597bc.
The commit wasn't fully correct as it was not installing properties
for counts in format-switching subtables.
Before, equivalent types, eg. otTables.BacktrackCoverage, etc,
where a subclass of their equivalent parent type, eg. otTables.Coverage
With this change, they are the same type.
The otBase and otConverters changes were needed to make the above
happen and still generate correct XML for SubTable type.
