Add new Opentype STAT table #96

tshinnic · 2017-05-17T02:23:34Z

New file src/tables/STAT.js will decode/encode Opentype STAT table data.
The undocumented struct version 0x00010000 is handled as example font files
included this version number. This new table type is added to those
mentioned in src/tables/index.js.

The test file executes using sample STAT tables from four different
open-licensed font files. Both input binary data and expected results
are included for each sample. Each sample is slightly different and
exercises variations on the STAT table format. Only the AxisValue
format 3 is not tested by these samples.

One additional open-licensed font file is included, Selawik-variable.ttf,
as mentioned in the test file. This font file demonstrates many of the
latest Opentype variation table additions. The test includes extracting
a STAT table from this font file, exercising decoding by mainline code.

The test file lastly tests encoding back to original binary data.

New file `src/tables/STAT.js` will decode/encode Opentype STAT table data. The undocumented struct version 0x00010000 is handled as example font files included this version number. This new table type is added to those mentioned in `src/tables/index.js`. The test file executes using sample STAT tables from four different open-licensed font files. Both input binary data and expected results are included for each sample. Each sample is slightly different and exercises variations on the STAT table format. Only the AxisValue format 3 is not tested by these samples. One additional open-licensed font file is included, Selawik-variable.ttf, as mentioned in the test file. This font file demonstrates many of the latest Opentype variation table additions. The test includes extracting a STAT table from this font file, exercising decoding by mainline code. The test file lastly tests encoding back to original binary data.

tshinnic · 2017-05-17T02:25:05Z

Any problems with source code formatting?

The new font file has many of the newer font variations OTF tables, including the STAT table, and so I wanted to include it in the pool of test fonts. It is OFL licensed in the original production font, and this variation includes that same OFL mention in its name table strings.

This font file includes tables:

    Listing table info for "test\data\Selawik-variable\Selawik-variable.ttf":
        tag     checksum   length   offset
        ----  ----------  -------  -------
        GDEF  0x8B19D41C    20324      460
        GPOS  0x8BC932CD    75998    20784
        GSUB  0xA8746161     1056    96784
   >    HVAR  0x03286591      985    97840
        OS/2  0x31823A5B       96    98828
   >    STAT  0xECB0D1A2      120    98924
   ?    TSI0  0xCC020E14     3112    99044
   ?    TSI1  0xE56E14B9   240060   102156
   ?    TSI2  0xCBE31AE0     3112   342216
   ?    TSI3  0xF0988744    68462   345328
   ?    TSI5  0x00C000C0      768   413792
   >    avar  0x1C2F2AAC       26   414560
        cmap  0x88858D10     1044   414588
   >    cvar  0x63AD1FA1       62   415632
        cvt   0x1EC01721      428   475660
        fpgm  0xD20A2786     4751   476088
   >    fvar  0x8B336992       76   415696
   >    gasp  0x0007001B       12   415772
        glyf  0x01FF8350    27072   415784
   >    gvar  0xE88CCDFB    26142   442856
        head  0x2B9471EA       54   469000
        hhea  0x2A3C24CC       36   469056
   >    hmtx  0x66714A29     1534   469092
   >    loca  0x58943F62      770   470628
        maxp  0x05CD1956       32   471400
        name  0x7C578249     1493   471432
        post  0x81071D38     2729   472928
        prep  0x0209EBE7      663   480840

and so this font could be useful for other testing/development.

Please let me know any other concerns, and I'd expect you'll have me do at least some changes... :)

devongovett

Thanks a lot for implementing this! Just a few formatting things.

devongovett · 2017-05-18T06:09:20Z

src/tables/STAT.js

@@ -0,0 +1,54 @@
+import r from 'restructure';
+
+


Extra whitespace line

Fixed in all places

devongovett · 2017-05-18T06:09:28Z

src/tables/STAT.js

+  axisTag:      new r.String(4),    // e.g. 'ital', 'wght'
+  axisNameID:   r.uint16,
+  axisOrdering: r.uint16,
+})


Missing semicolons

Oh frabjous day. Attempting to check for missing semicolons also in the test file, to copy styles seen in other test files, eslint just kinda mentioned that I wasn't actually using a couple of the binary data sample tables 02, 03, 04. Missing semicolons and missed tests verified. Other cleanups done.

devongovett · 2017-05-18T06:09:45Z

src/tables/STAT.js

+  },
+  // Obsoleted version 1.0 ?
+  0x00010000: {
+  },


Put this on the same line as the opening brace

devongovett · 2017-05-18T06:11:12Z

src/tables/STAT.js

+    valueNameID:    r.uint16,
+  },
+  1: {
+    value:          r.fixed32be,


Everything is assumed to be big endian by default. r.fixed32 should work and is an alias for r.fixed32be

devongovett · 2017-05-18T06:14:05Z

test/tableSTAT.js

+  0x00, 0x00,                       // 0004     uint16  flags
+  0x01, 0x05,                       // 0006     uint16  valueNameID
+  0x02, 0xbc, 0x00, 0x00            // 0008     Fixed(16.16)  value  700.0
+];                                  // 0078


Instead of including the binary data in the source code, can we just include the font file in the repo and read the table directly in the tests?

As mentioned in change message and test file, checking even the few example font files I could find with STAT table, found differing data formats. Interestingly the deprecated/undocumented STAT format 1.0 was found twice. My intent with the sets of binary/expanded data was to exercise the format variations found so far, including missing or present (non-null) data. Otherwise the older format 0 file would have been a surprise requiring a fix.

I was thinking including the self-contained binary data in the test was better than committing to including multiple font files. And certainly thinking that testing variations is better than not.

If Selawik-variable.ttf is included as proposed here, then one of the binary stanzas can go away in favor of reading from the file directly. (I picked this font file to include because of all the other newer tables it also contains)

The TestGVAREight.ttf is from the same series of test file you have already sampled, and is quite small. We could switch to include it here and skip that binary table.

Questions:

multiple tests where variations seen? (formats, data presence/absence, etc.)

self-contained data for multiple testing where sample fonts are otherwise uninteresting?

what guidance on when to include new font files to drive testing?

Notes:

481,504 Selawik-variable.ttf
106,184 AmstelvarAlpha-VF.ttf
264,380 Decovar-VF.ttf
4,692 TestGVAREight.ttf (from text-rendering-tests)

devongovett · 2017-05-18T06:16:50Z

test/data/Selawik-variable/Selawik-README.md

+# Hinting
+This version of Selawik is primarily hinted with the light Latin hinting style Microsoft recommends for variable Latin fonts. The VTT Light Latin autohinter was used to create the first round of hints, which were then reviewed and touched up. 
+
+This hinting style only uses CVTs for vertical metrics anchors (ascender, descender, cap height, x-height, and baseline). While this makes for an easy job hinting a Latin font, it makes for a poor test case because Selawik doesn't vary vertical metrics with weight, thus doesn't vary CVTs, thus doesn't need a cvar. So, to make it more interesting, we added CVT-based stem hints to the lowercase only. This provided the need to vary CVTs and thus require a cvar. Note that this was only done for testing purposes. For variable fonts, Microsoft recommends the light hinting style using the `ResYDist()` command instead of a CVT-based stem hint. 


Is there a license available for this font that we can include?

Both the original selawk.ttf and this modified Selawik-variable.ttf have the same notices in the name table:

0: © 2015 Microsoft Corporation (www.microsoft.com), with Reserved Font Name Selawik. Selawik is a trademark of Microsoft Corporation in the United States and/or other countries. 13: This Font Software is licensed under the SIL Open Font License, Version 1.1. 14: http://opensource.org/licenses/OFL-1.1

I had ginned up an OFL.txt that began with these lines:

Copyright (c) 2015 Microsoft Corporation (www.microsoft.com), with Reserved Font Name Selawik. Selawik is a trademark of Microsoft Corporation in the United States and/or other countries.

This Font Software is licensed under the SIL Open Font License, Version 1.1.
This license is copied below, and is also available with a FAQ at:
http://scripts.sil.org/OFL
followed by the OFL 1.1 text.

While I can't find a copy of any license file in the (original 2015) Selawik_Release.zip, I kinda felt presumptuous generating the missing license file. I'll do so to match your other font file entries if you wish.

devongovett · 2017-05-18T06:18:49Z

test/tableSTAT.js

+
+    let readEncoded = estream.read()
+    // Is there a better way to extract bytes from stream buffer?
+    let readArray = readEncoded.toString('binary').split('').map((c) => c.charCodeAt())


You can do Array.from(readEncoded)

devongovett · 2017-05-18T06:19:55Z

test/tableSTAT.js

+//  let result = tableSTAT.decode(stream);
+    let result = tables.STAT.decode(stream);
+
+    //console.log('    decoded STAT table: %j', result);


Remove the commented logs

devongovett requested changes May 18, 2017

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add new Opentype STAT table #96

Add new Opentype STAT table #96

tshinnic commented May 17, 2017

tshinnic commented May 17, 2017

devongovett left a comment

devongovett May 18, 2017

tshinnic May 28, 2017

devongovett May 18, 2017

tshinnic May 25, 2017

devongovett May 18, 2017

devongovett May 18, 2017

devongovett May 18, 2017

tshinnic May 25, 2017

devongovett May 18, 2017

tshinnic May 25, 2017

devongovett May 18, 2017

devongovett May 18, 2017

Add new Opentype STAT table #96

Are you sure you want to change the base?

Add new Opentype STAT table #96

Conversation

tshinnic commented May 17, 2017

tshinnic commented May 17, 2017

devongovett left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment