MuseJazz Font is missing european diacritical characters

Not only German Umlauts, also french accent letters (e.g. à, ê etc) as well as other special characters like µ and €.

Well, french and czech would be needed...

Indeed, MuseJazz only has the basic ASCII character set. If anyone who has experience extending with the various diacritics (?) wishes to help, that would be greatly appreciated. I gather there are semi-automatic ways of generating these characters. Note there are actually dozens, maybe hundreds of characters that could be added, based on what other fonts provide. There are additional characters used in Spanish, Portuguese, French, and other languages in addition to German, plus of course the non-Latin languages.

I gather there are semi-automatic ways of generating these characters.
Could you point to any website with instructions?

It was in the Font Forge documemtation somewhere, that's about all I remember.

Maybe this page ?

I get the sense it's still a bit of work doing that, at least when consider the sheer number of such characters needed when all languages are considered.

However, it's possible we could pull this off more collaboratively. If a few people with experience in one language other English languages can identify the necessary characters for their language, and then maybe come up to speed on the process of adding characters to a font (it;s not *that* complicated), maybe we could at least knock out the most important characters sooner rather than later? On Windows, at least, MuseScore will use a locally-installed version of MuseJazz in preference to the compiled-in version, so the reward for contributing would be the ability to use the new characters instantly, at least in your own prints and PDF's if not for scores shared with others in MSCZ format.

I'd be happy to collect the various versions created by others and copy the glyphs back into the main source if necessary. i gather that in the past, the font binary was used as the source, making automated management via SVN difficult, but but Werner has been encouraging use of SFD sources, and I do have an SFD for MuseJazz in its current form.

I'd want to tweak the kerning before anyone gets started, though.

I could do French and Spanish. Could you link to a working binary of fontforge?

French can use Œ and œ as well and spanish reversed marks ¡ and ¿

FontForge is inherently a Linux program; there is no native Windows or Mac version as far as I know. But there a package called fontforge-mingw that runs on Windows using some sort of Linux emulation mode or something, and that's what I've been using. As far as I can tell, the best place to get it from is http://www.geocities.jp/meir000/fontforge/. You just have to ignore the Japanese characters and scroll down to find the mingw version, which is supposed to be easier to onstall and run than the cygwin version.

I mentioned the kerning because Jörg brought it up. I did add some kerning for 1.1, but not as much as I could have as it took me forever to figure out how to get it working at all. There are definitely still some problematic letter combos where it looks like there is too much space between them. Now that I have a better idea of what I'm doing, I'd like to improve the kerning of the existing characters before we add new ones, on the theory that this might make it easier to set things up so that the new characters get kerned similarly to the ones they are based on. I know FontForge supports the idea of kerning classes, which are sets of characters that should kern the same. But I don't really know for sure how that works. It might be just as easy to adjust the kerning later, and actually, I can't be sure that "a" and "ä" really should kern the same. But it sure can't hurt to look at kerning first, so let me do that this week.

My understanding is that kerning is handling by the Qt libraries, and there is little MuseScore can do about it. I don't know if 1.1 is built with the latest Qt libraries or not; I do seem to recall something about Qt having improved kerning at some point?

Anyhow, when I was doing the kerning originally, I was having trouble getting MuseScore to pay attention to it at all so I did most of my testing in Wordpad and LibreOffice. Eventually I found an option I needed to set in FontForge to get the kerning to work with MuseScore. It's entirely possible I still didn't do this right. But since your example shows differences even with Times, I guess I won't worry about that for now.

Anyhow, I still figure it makes sense to get the kerning as good as I can, probably continuing to use Wordpad & LibreOffice but also checking in MuseScore. So I'm still aiming for end of week.

@jorgk3 Which version of MuseScore did you use for your test? For the linux one, which Qt version? Which windows version?
I tried to write the same sentence in MuseScore 1.1 (Qt 4.7.3), windows vista and I don't have the same result at all on screen, on PDF or on PNG.

MuseScore, PNG, PDF

Reinstating original title

Hmmm... in lasconic samples, I would say the main point is perhaps not kerning:

Between the first and third sample the fonts seem different: the copyright symbol is elliptical in 1st and round in 3rd; the top grace of 'i' is more slanted in 1st than in 3rd; 'e' is narrower in 1st then in 3rd...

In general, the design of the font in 1st sample looks more condensed than the font in the 3rd sample.

My impression is that in 1st sample, some font substitution occurred: the app seems to 'believe' to be using the font actually used in 3rd sample and keeps the same metrics but it is actually using another font with a more condensed design, whence the 'airy' look around the individual glyphs.

This may look like a kerning issue but, if font substitution occurred, may have nothing to do with kerning at all, but rather with differences in font style.

M.

The issues with how MuseScore kerns, or whether it is sometimes doing a substitution, aside, it's definitely the case that Musejazz was in need of more thorough kerning. I've made another pass over it, looking at all letter combinations (thanks, Joerg, for the sample text!). I probably overdid it. Or more to the point, I probably should have looked at the padding around some the characters themselves rather than doing everything via kerning (I did widen a few characters' boxes). But anyhow, the deed is done, and I think it looks a lot better. There are still a couple of letters, like "r", "t", and "f", where the shape of the letter is such that it still can look weird around them. I improved "r" slightly in this respect (angled it up more). I now leave it to anyone who wants to add more characters with diacritics. You can either take turns passing it around, or all pass my your versions back and I'll collect the new characters. If anyone wants to take another look at kerning though, we will have to be careful not to step on each other there.

I've attached the SFD file and the generated TTF via a ZIP file.

Oh, one other things I keep forgetting to mentioN:

On my systems (Vista), I can make MuseJazz bold, and it looks pretty decent. I didn't actually create a seaprate bold font, but whatever Windows and/or Qt do to fake it, seems to work OK. So I used bold title in the Jazz Lead Sheet template. But the images and PDF files generated on musescore.com don't do that, so I'm guessing that server is running Linux, and the fake bold thing doesn't work on Linux. Or there is some thing that needs to be set set in FontForge to allow that. Anyone know?

Definitely. Go ahead. Same for the bold issue. Regarding diacritical glyphs, I let you go ahead.

The recommended way to create characters with diacritics is through reference (or composition):

1) create the diacritic(s) themselves as independent glyphs, possibly in their proper Unicode points, paying attention to the fact that, for many diacritics, Unicode distinguishes between spacing and non-spacing diacritics: the formers are actually independent characters, the latters are 'real' diacritics (usually the Unicode name is "COMBINING ...").

2) create the composite glyph using a reference to the base character and a reference to the diacritic.

For instance, rather than creating uumlaut by copying the outline of 'u' and adding to it two dots, create the diaeresis diacritic (the proper point is U+0308) then create uumlaut (U+00FC) with a reference to 'u' and a reference to diaeresis.

Using FontForge, a reference to a glyph is 'copied' by selecting the glyph in the font window and pressing Ctrl-G, and then 'pasted' by selecting the destination glyph and pressing Ctrl-Shift-V.

This has several advantages: the font itself is smaller, the rasterizer is somehow faster (it may not need to rasterize the 'u' part if it has already rasterized the 'u' itself) and any hint, anchor, etc... of the individual parts are immediately available into the composite glyph.

Additionally, when the reference to a base char (for instance 'u') is pasted into an empty glyph, FontForge also copies the metrics ensuring some consistency. If the individual parts already exist, FontForge is able to automatically create a good deal of the known composites, by using the relevant references.

Strictly speaking, the reference way should not be used for composite glyphs whose parts overlap (for instance 'ç'): TT documentation used to warn against overlapping referenced outlines; however, many fonts from reputed foundries use references also in this case and no font engine I know actually has problems with this.

Hoping it may be useful...

M.

So, I'm looking at MuseJazz again with hopes of submitting improvements in time for a 1.2 release, whenever that turns out to be. I had considered starting over with the character spacing and redoing the kerning again, plus Scribus was having problems with the font, and I just wanted a fresh start. But on further examination, it's the irregular shapes of the characters that leads to the need for so much kerning, and it turns out the most recent version of Scribus works just fine with MuseJazz,

So, the SFD file I posted still makes as good a starting point as any for adding characters. So, who is on board for helping out with adding characters? Thanks, Miwarre, for the info on how to accomplish the diacritics. I understand enough about German to handle the umlauts, and I can draw a freehand ß that probably won't look any worse than most, but that's all I'd feel up for.

One natural division would be to say one person per language to come up with the complete list of new glyphs needed (and to test the final results in that language), one or more people to add the necessary outlines for the umlauts and accents and any glyphs that cannot be created by combining through reference, and one person to do the work of creating the combining characters by reference. Of course, any one given person might take on more than one of these tasks.

Of these, the only thing I know how to do already is add outlines. Although I don't actually speak any language but English (and whatever I remember from a couple of years of German in school), I can probably do as good a job of drawing outlines given a picture of what I am trying to draw as anyone else who is not already skilled at this. So I would be willing to do any or all outlines. I"m also perfectly willing to learn the whole combining-by-reference process (sounds straightforward enough) if no one else wants to take that on.

BTW, one language I'd like to see covered that no one else has mentioned is Portuguese.

Sounds great! You take ß, and I'll endeavor to make sure we've got all the glyphs needed for German, French, Spanish, and Portuguese, based on the links given here and whatever other info I can find. If there are any other glyphs (like perhaps the "c," or "oe" characters) that really need to be designed from scratch, I may hold off on those for now in hopes someone else may step forward. And hopefully some native speakers can give it a once-over when we're done. As for Czech or any non-:Latin language anyone else would like to see, I would say, who ever asks for it just volunteered to do the legwork :-)

I should be able to have my part done by the end of the week if I don't run into any unexpected technical hurdles.

Dutch: ij
Scandi: å
Spanish: ¿, ¡
Czech: š i.e. an invers ^ on top of an s

It turns out that creating accented characters is surprisingly easy. In addition to the manual method of building a character from references that Miwarre describes, FontForge provides a more automated method, where it automatically locates and combines references to the necessary characters, and then all I have to do is drag things around as necessary. I've also added the upside down exclamation point and question mark.

But I'm not sure about the best way to to deal with the fact that accented capital letters are taller than ordinary capital letters. By default, FontForge apparently does something to increase line spacing to make room for the accents, even though the values given for ascent, descent, and em size haven't changed. These vertical metrics are something I struggled with for some time to get right for 1.1, but it was more by luck that I got it sorted out. I am having trouble finding references online to how this is supposed to work in the presence of accented characters. Any suggestions or pointers would be most appreciated!

Ligatures (combinations like "oe" squished together) are also still a bit mysterious to me. I can see the basic process for creating them, but I don't understand if they are thing that MuseScore would have a chance of "automatically" recognizing and transforming, the way, say, it automatically turns "(c)" into the copyright symbol, and if so, if it's possible to make that language-dependent.

Looks like I kind of blew it when I extended MuseJazz for 1.1, in that I didn't leave room for accents above capital letters. No easy way to fix it without messing things up in other ways. I have elected to shrink the capital accented letters, so that even with accent they are no taller than unaccented letters. Not ideal, obviously, but my hope is that people won't notice (unless they do something in all caps with mixed accented and unaccented letters). I have also gone ahead added an eszett because I was on a roll, but feel free to redo it. I actually added all the accented characters and ligatures from the F2 text symbol palette (except the fractions).

The font has been temporarily renamed MuseJazzEuro so I could test it side by side with the original version. I plan to rename it back before submitting it. But if anyone wants to try it out now and give feedback, here it is.

Yeah I try it :) Thank you
Bernhard

Thanks for the suggestions. I know the eszett is big and ugly; that's how it usually looks to me :-). I don't remember any email from you with suggestions on the design of the characters - was it recent? Do you think the Times version makes a good model?

I'll take a look at your other suggestions. In some cases, I deliberately pushed an accent character out of what I thought of as its natural position in order to not exceed the line height; the alternative was to shrink the letter even further. But I can try to tweak those. Not being a native speaker of any language that uses such characters, I have no innate sense of how "off" these look. So thanks again for the feedback.

Thanks for the heads up - I just found your messages in my spam folder. Not sure why; gmail usually has a pretty good hit rate. I'll trying to figure out how tell gmail you are approved. Hopefully adding you to my address book will help.

Anyhow, the reason horizontal centering affects height is that when centered a large capital letter, an accent might exceed the lone height. Moving it off center allows me to lower it a bit. But I can certainly revisit those characters.

As for eszett - are you already versed in FontForge, or do you jave some other outline editor you could use? If so, great! If not, contact me offline (assuming it works this time) and I'll give you some pointers.

Tweaks based on Jörg's suggestions:

Thanks! As I indicated via email, I don't want to deprive you of the joy of glyph design, and I am at least half serious about that. I figured I would give it one more shot. If it's good enough as is and you're not feeling like spending time on this, great. But I would also be perfectly happy if you were to replace it with something you thought better.

The thing I don't have a sense of, not having ever drawn hundreds of these things freehand in grade school penmanship classes, is what the "rules" are for it. In particular, what is supposed to distinguish it from a capital B. I can always tell the difference when I see it, but I have never been taught exactly what is going on. So my first attempt erred on the side of ugliness - it's clearly not a B, because a B wouldn't look that strange :-). The new version looks very much like a B to me, but not the B in this particular font. Is that enough? I don't have the context to answer that.

Here is the ZIP with the SFD and TTF files. I think there may have one one more very small tweak since that screen shot, but I can't remember which character. I think maybe I reduced the size of the accent on the i-grave to make it less likely to collide with other characters. I have added only a tiny bit of kerning info for these new characters - for the most part, they just inherit the kerning of their parents. Meaning the accents can possibly now crowd other characters too much, which is a good reason to limit the horizontal size of the accents, and to keep them as close as practical to the letters vertically. People who actually use these letters regularly might want to play with them and see if there are any obvious problems with any of these letters crashing into other letters when used together.

Thanks for all your hard work, Jörg! I added the IJ and ij for Dutch, but I'm thinking we're done for now. If someone wants to design the characters needed for some language, feel free to take a crack at it if you know your way around FontForge - but try not to break anything!

Heads up for anyone playing with this: I just added glyphs for the solfege syllables (starting at 0xe201), for use in chord symbols in Italy and any other countries in which that is traditional. Do, Re, Mi, Fa, Sol, La, Si, plus Ut, So, and Ti. Here are the updated files.

The MuseJazz font in the SVN repository is 6 months old (both the branch and the trunk version).

Would it not be the case to update it with the latest version, so that it can be shared easily?

Thanks,

M.

I don't think I do either. I have installed Tortoise SVN, but haven't tried doing anything with it yet. Back when I first started working on this font for 1.1, I was the only one, and it seemed simpler for me to just email my updates to someone who already had access. Plus, I didn't (and still don't) have a build environment set up to test any changes I made in context. Long term, I know that's not the right way to go. But for the record, I did send this most recent version in yesterday just after posting it here.

I commited the fonts and the chord description files in r5277. I tested a bit and I missed the € sign and the £.
Moreover, the ï is different from the other ones (ë, ä, ö, ü)

The @ is also quite weak.

Some more ideas for the @ and other characters we might want in future versions.
See JazzText in http://www.jazzfont.com/JFDemo.pdf

Please add missing Lithuanian letters: ĄąČčĘęĮįŠšŲųŽž

Don’nt forget esperanto Ĉ Ĝ Ĥ Ĵ Ŝ Ŭ

Btw, why Musescore does not just rely on system «Special characters» table .and locale

Presently my system gives direct access to any latin european character with diacritic, but they are not received by the text being typed in musescore. I have to copy/paste them...

I don't know what "system" table you refer to - remember, MuseScore runs on several different systems. If you are experiencing oroblems on your particular operating system, with your particular keyboard, then you might want to open a discussion in the Support forum - there might some simple solution. Or, it might just be a limitation in the Qt framework upon whivh MuseScore is built.

But whatever it is, presumably it has nothing to do with the MuseJazz font in particular.

I don't know what "system" table you refer to - remember, MuseScore runs on several different systems. If you are experiencing oroblems on your particular operating system, with your particular keyboard, then you might want to open a discussion in the Support forum - there might some simple solution. Or, it might just be a limitation in the Qt framework upon whivh MuseScore is built.

But whatever it is, presumably it has nothing to do with the MuseJazz font in particular.

Any system (Linux, Mac, Windows) give access to any UTF 8 character, (keyboard or table… in my french Ubuntu Accessoires>Table de caratères)
Of course I’ll get the useful character only if the used font provides the glyph.

So I think that (whichever the font) any UTF 8 character should be typed the same way as it is when using a terminal (kbd or characters table). I don’t understand why we shall use this «F2» window, where the quest for glyphs seems different of the one in the characters table of my system

If you have rather that I open a new thread. just let me know.

This thread is becoming too long and the last posts doesn't have anything to do with MuseJazz. I propose to close it and open a feature requests for the missing characters and a forum discussion for whatever MuseScore does badly.

In fact, I'm not proposing :)

#15556: Add eastern europe characters to MuseJazz
#15557: Add € and £ to MuseJazz, review @ character

I let you do the forum discussion : http://musescore.org/en/forum