Working on way to subset fonts for ePub/KF3

[pdurrant]

Originally Posted by JSWolf

But is there a way of taking a TTF or OTF font file and subsetting it that works on Windows, OS X, & Linux without the need for Cygwin?

That is rather what we're looking at in this thread. There seems to be code, even python code, out that there might do the job. It's a question of someone having the time to look into it properly, and adapt it for this particular purpose.

Or finding that someone else has already done it.

[Doitsu]

Originally Posted by JSWolf

But is there a way of taking a TTF or OTF font file and subsetting it that works on Windows, OS X, & Linux without the need for Cygwin?

This is yet another post that clearly shows that don't really understand what you're talking about. Cygwin was primarily created to enable Unix shell programming under Windows. Both Linux and OS X are UNIX derivatives and therefore don't require Cygwin. (Both also natively support the most common scripting languages and UNIX shell programming.)
I.e. Cygwin is not the problem. The problem is adapting existing scripts for this special task.

@pdurrant:
I had a look at the links that you posted and it seems that the first one fits the bill perfectly. The Web Font Optimizer web site uses it for subsetting and it works nicely with the predefined free fonts, but it's not possible to select custom fonts.
I tried to create an offline version by pulling the source files, but I couldn't get it to work, because there was one dependency that I couldn't resolve.
(It was probably just a stupid beginner's mistake like a wrong folder structure or a missing or outdated CPAN module.)
Maybe someone with better technical skills than me could get it to work offline.

[tobias.wantzen]

Any progress on this topic meanwhile? It's still very interesting I think.

Have you thought about this tool:
http://www.fontsquirrel.com/tools/webfont-generator

In the Export settings you can choose »Custom Subsetting« in any way you want.

Cheers
Tobias

[Freeshadow]

Kovid has an implementation inside calibre and Toxaris works on a stand-alone. Both worth mentioning. Search the forum.

An 3rd party online tool like squirrel isn't the optimal solutions unless you feel good with sending your files to somebody's servers.
Personally I won't see the problem fully solved until I see a tool fully capable of parsing epub and trimming every font file inside while respecting ligatures and used alternative chars. As for now this is (like many others) an optimisation step requiring manual tweaking.

[Hitch]

Originally Posted by Freeshadow

Kovid has an implementation inside calibre and Toxaris works on a stand-alone. Both worth mentioning. Search the forum.

An 3rd party online tool like squirrel isn't the optimal solutions unless you feel good with sending your files to somebody's servers.
Personally I won't see the problem fully solved until I see a tool fully capable of parsing epub and trimming every font file inside while respecting ligatures and used alternative chars. As for now this is (like many others) an optimisation step requiring manual tweaking.

I've used Toxaris' method a number of times without any problems. Yes, it has a few manual steps, mostly tedious; (open the ePUB with whatever unzipping method you want to use, copy the font file(s) over somewhere else, merge the files, copy the text/html, put it in Toxie's box, click for the subset, then make the new subsetted font(s).) Then replace all the original fonts with the subsets. It's a bit tiresome, but it works, which is a huge help.

I will say, though, that Barb tells me that it crashed her computer. I think that it's simply not that intuitive to use; it's not clear that you do a-b-c (the sequence, I mean). I know that sounds idiotic, but it isn't. If I'd written it myself, I swear I'd have put numbers next to the various fields. I have to look at it every time I use it to remind myself "paste the text here first, then get the subset, then put the original font file in there, then click that, then..." As I said, yes, I know it makes me sound idiotic, but I don't find it that intuitive in terms of sequence, and I suspect that something along those lines is what caused her computer to crash. (To be precise, she had to manually reboot it).

Maybe it's as simple as the "select font" and "New Font" fields are above the Characters field, when you need to determine the characters first. {shrug}. And the Create subset button is above the "Characters in Subset" box; I think I actually clicked that once early on, thinking it was the "determine unique characters" function; don't know precisely what it is, but it does throw me. I have, in a time crunch, used Calibre's function and swapped out subsetted font sets, but I prefer to use Toxaris' macro, as it never feels quite right to me to use Calibre for anything on a commercial book, even something as small as subsetting.

FWIW.

Hitch

[Toxaris]

Hitch, I wasn't aware but it makes sense. For me the interface was logical, probably because I created it... I will release an intermediate version with numbers and a slightly changed interface.
The newer version that can open ePUB has been delayed due to other obligations. The OS/2 Version change will not happen for now. It seems that is too hard to do for me other than utilizing FontForge.

[Hitch]

Originally Posted by Toxaris

Hitch, I wasn't aware but it makes sense. For me the interface was logical, probably because I created it... I will release an intermediate version with numbers and a slightly changed interface.
The newer version that can open ePUB has been delayed due to other obligations. The OS/2 Version change will not happen for now. It seems that is too hard to do for me other than utilizing FontForge.

Oh, Tox:

I'm sorry. I wasn't complaining, that's not what I meant. I was trying to really say something positive about it, but in reviewing it, I see it came out very negative-sounding, sorry. But I admit, I do have to think about it every time I use it, just because, being the OCD-ness that I am, I expect "stuff" to go from top-to-bottom. Dumb, dumb, dumb, I know. Don't take my dumb-ness as meaning that anyone else finds it confusing, please. ;-) It's a super tool and I'm happy to have it.

Hitch

[Toxaris]

Don't worry, I know you by now. You had valid points. The new version I posted works from top to down...

[JSWolf]

Originally Posted by Freeshadow

Kovid has an implementation inside calibre and Toxaris works on a stand-alone. Both worth mentioning. Search the forum.

An 3rd party online tool like squirrel isn't the optimal solutions unless you feel good with sending your files to somebody's servers.
Personally I won't see the problem fully solved until I see a tool fully capable of parsing epub and trimming every font file inside while respecting ligatures and used alternative chars. As for now this is (like many others) an optimisation step requiring manual tweaking.

The Calibre method works. Why not give that a try. It also deletes unused font files. You can use the new Polish books feature and just subset the fonts.

[Hitch]

Originally Posted by JSWolf

The Calibre method works. Why not give that a try. It also deletes unused font files. You can use the new Polish books feature and just subset the fonts.

Wolfie:

I can't quite explain it. {shrug}. Somehow, I just don't feel right about it. It's inexplicable, from a logic standpoint. I know it does not make a difference, but that could drive a person batty. Just bad juju, I guess. I mean...we get asked, at least 5x a week, two questions, those being:

1. How do I know you won't steal my book? (phrased less bluntly, but...), and,
2. Are you using Calibre to make my book?

...to which I answer "no" to the latter and try not to strangle someone for the former. (No, we don't have TIME to read these to even know which would be worth stealing, if we were dishonest!) The stealing question always gets my hackles up, but I understand why they ask the first question--they want to know that they're getting what they pay for, hand-crafted ePUBs, basically.

Hell. I don't know, it certainly would be fast. ;-)

Hitch

[tobias.wantzen]

Thanks for your answers. I hadn't seen Toxaris’ tool. To save a bit of search work for all those who stumble across this thread - FontShrinker could be found here:
https://www.mobileread.com/forums/sho...d.php?t=203163

Tobias

[JSWolf]

Originally Posted by Hitch

Wolfie:

I can't quite explain it. {shrug}. Somehow, I just don't feel right about it. It's inexplicable, from a logic standpoint. I know it does not make a difference, but that could drive a person batty. Just bad juju, I guess. I mean...we get asked, at least 5x a week, two questions, those being:

1. How do I know you won't steal my book? (phrased less bluntly, but...), and,
2. Are you using Calibre to make my book?

...to which I answer "no" to the latter and try not to strangle someone for the former. (No, we don't have TIME to read these to even know which would be worth stealing, if we were dishonest!) The stealing question always gets my hackles up, but I understand why they ask the first question--they want to know that they're getting what they pay for, hand-crafted ePUBs, basically.

Hell. I don't know, it certainly would be fast. ;-)

Hitch

To me it makes no sense. I've been using the font subsetting in Calibre and it works well. I don't see any reason not to use it. The answer to #2 would be... "We are just using Calibre to subset the embedded fonts. Calibre is not used for any sort of conversion."

[Hitch]

Originally Posted by JSWolf

To me it makes no sense. I've been using the font subsetting in Calibre and it works well. I don't see any reason not to use it. The answer to #2 would be... "We are just using Calibre to subset the embedded fonts. Calibre is not used for any sort of conversion."

Wolfie, my sweet:

I didn't say it made sense. ;-) In fact, I think I came right out and said it didn't. What can I say? Chalk it up to my own eccentricity. After all, I wouldn't be the only eccentric here, now, would I?

Hitch

[Toxaris]

To me, it does make sense though.

[Turtle91]

Makes sense to me as well. Trying to explain what font subsetting IS would be hard enough, let alone trying to explain how you only use Calibre for one thing and not another. The only thing the customer would hear is "yes, we are using Calibre to convert your book but charging you a 'hand crafted' price."