| #!/usr/bin/env python3 |
| |
| import sys |
| import array |
| from gi.repository import HarfBuzz as hb |
| from gi.repository import GLib |
| |
| fontdata = open (sys.argv[1], 'rb').read () |
| text = sys.argv[2] |
| # Need to create GLib.Bytes explicitly until this bug is fixed: |
| # https://bugzilla.gnome.org/show_bug.cgi?id=729541 |
| blob = hb.glib_blob_create (GLib.Bytes.new (fontdata)) |
| face = hb.face_create (blob, 0) |
| del blob |
| font = hb.font_create (face) |
| upem = hb.face_get_upem (face) |
| del face |
| hb.font_set_scale (font, upem, upem) |
| #hb.ft_font_set_funcs (font) |
| hb.ot_font_set_funcs (font) |
| |
| buf = hb.buffer_create () |
| class Debugger (object): |
| def message (self, buf, font, msg, data, _x_what_is_this): |
| print (msg) |
| return True |
| debugger = Debugger () |
| hb.buffer_set_message_func (buf, debugger.message, 1, 0) |
| |
| ## |
| ## Add text to buffer |
| ## |
| # |
| # See https://github.com/harfbuzz/harfbuzz/pull/271 |
| # |
| # If you do not care about cluster values reflecting Python |
| # string indices, then this is quickest way to add text to |
| # buffer: |
| # hb.buffer_add_utf8 (buf, text.encode('utf-8'), 0, -1) |
| # Otherwise, then following handles both narrow and wide |
| # Python builds (the first item in the array is BOM, so we skip it): |
| if sys.maxunicode == 0x10FFFF: |
| hb.buffer_add_utf32 (buf, array.array ('I', text.encode ('utf-32'))[1:], 0, -1) |
| else: |
| hb.buffer_add_utf16 (buf, array.array ('H', text.encode ('utf-16'))[1:], 0, -1) |
| |
| |
| hb.buffer_guess_segment_properties (buf) |
| |
| hb.shape (font, buf, []) |
| del font |
| |
| infos = hb.buffer_get_glyph_infos (buf) |
| positions = hb.buffer_get_glyph_positions (buf) |
| |
| for info, pos in zip (infos, positions): |
| gid = info.codepoint |
| cluster = info.cluster |
| x_advance = pos.x_advance |
| x_offset = pos.x_offset |
| y_offset = pos.y_offset |
| |
| print ("gid%d=%d@%d,%d+%d" % (gid, cluster, x_advance, x_offset, y_offset)) |