While dabbling on FanDoc support for tcolar's NetBeans plugin, I noticed the hyperlink notation. I don't know if this symbol is particular easy to find on English/American keyboards, but it's a tremendous hassle on Scandinavian keyboards so I was wondering if this has been considered and whether there's a chance of another, less exotic character instead?

For instance these two: SQUARE_BRACKET_START url SQUARE_BRACKET_END SQUARE_BRACKET_START name PIPE url SQUARE_BRACKET_END

The square brackets and pipe character international developers are already familiar with.

Actually the syntax you propose is exactly what is use on my own wiki (which my site runs on).

Anyway, in Fan any URI(not just in fandocs) are defined using that char(back-tick), so I suppose if it's not avail on foreign keyboards that might be a pain even when writing code.

Maybe there could be another alternate syntax for defining URIs?

Yeah I believe even our wiki at work (Confluence) uses that syntax. It's late and dark here which doesn't help, but I have yet to actually find the back-tick on my keyboard! lol

Edit: My girlfriends French/Canadian keyboard has it (right next to <Enter>), but that's presumably only because it's used for French accents?!

AFAIK on Scandinavian keyboards the back-tick and the fore-tick (something most English keyboards lack) is left to the backspace, but is a bit tricky. Pressing the key alone doesn't do anything. But if you press it (not hold it) followed by pressing space it should give you a fore-tick ´, pressing it while holding Shift and than a space should give you a back-tick `. And you should be able to write most letters with accents by pressing the key followed by an a, e or whatever, like á, à, é, è, ...

`'´

Off-topic rant: it sucks when some people combine a back-tick with an apostrophe like `some text'. That doesn't make any sense. Use two apostrophse, combine a fore-tick and a back-tick or use LEFT SINGLE QUOTATION MARK (U+2018) and RIGHT SINGLE QUOTATION MARK (U+2019) which probably would be the best solution, but don't quote with back-tick and apostroph.

BTW, I used three back-ticks to wirte them in this post as I got an "Invalid URL" error, otherwise.

Spot on helium, now it's bright morning and I can actually see the keyboard, the back-tick is indeed next to the backspace. As you correctly mentions though, it is not directly usable on Windows (and isn't even mapped at all on the Ubuntu 9.04 I'm on now).

Its already been mentioned, but the Fandoc syntax reuses the same URI literal syntax as the language itself. So I think that is the basic problem.

I am pretty much using every character available on a US keyboard. Just about all those chars are also commonly used in Java with the exception of ` backtick and # pound chars.

Are those chars a problem for anybody using non-US keyboards? If so should we consider an alternate syntax to make it easier to use those chars?

Well I had completely missed it being literal syntax too, that makes the problem even bigger. I can see how it makes sense on a US keyboard.

I definitely think this needs to be addressed if you wish to capture any kind of international crowd. I guess there's a reason why it is not part of the usual alphabet of other languages.

Hash/pound is a non-issue however, it's no more exotic on the keyboards I know (Scandinavian and German) than dollar, ampersand etc.

I definitely think this needs to be addressed if you wish to capture any kind of international crowd. I guess there's a reason why it is not part of the usual alphabet of other languages.

One solution is allow two chars to be used for a Uri literal - the backtick for US keyboards, then select another char that is likely to be on non-US keyboards. Can you suggest one?

Another solution would be to use normal quotes with a prefix char:

u"http://fandev.org"
u'http://fandev.org'

Not sure I'm a fan (pun intended) of multiple ways, that would get annoying when doing any kind of collaborate work on code bases. I don't know, it just feels wrong... there should be a lowest common denominator.

Now I know how, I guess I could get used to hitting <Shift> + <Backtick> twice in order to write one single backtick. But it seems possible other language layouts would hide it even better or perhaps entirely. The fact that few (Perl?) programming language uses it as a token would suggest this to be the case.

And I'm afraid I don't quite know enough about the existing Fan grammar and international keyboards to suggest a good alternative, but perhaps wikipedia's keyboard layout page can form a base for some research: http://en.wikipedia.org/wiki/Keyboard_layout

Normal Unix shells (well, at least Bash), use the backtick for expanding command results. Python uses it for a similar effect, but I forget the exact meaning and I don't think it's commonly used there. LaTeX uses backticks to indicate leading quotation chars in the fashion complained about by Helium.

I agree with Casper that it should be the same for multiple human languages to allow better collaboration (and to avoid surprises).

Anyone familiar with traditional Python would expect u"..." to mean a Unicode string. Something to keep in mind, at least.

After surfing around, it seems the back-tick is used in a bunch of existing languages like Bash, Python, and Ruby. I know I've used it a quite a bit in Ruby for shell scripting. So it can't be that bad.

So I'm inclined to say we should leave things alone. But then again I don't want to be a self centered American :-)

What do others think? Is this enough of a problem to warrant a grammar change to the language?

Many programming style guides suggest using english names (for readability in the wider world) so I don't see how one character is any worse.

The issue has nothing to do with English names though, in fact it's common practice for international users to only use English characters for comments and identifiers, since many tools would choke on special characters anyway (i.e. the Danish å, ø and å).

I don't see that as orthogonal to the issue of exotic alphabet/tokens which can be difficult to enter - I'm a programmer from 8 to 16, I'd like the alphabet of my programming language to be easy to enter. Again, no general purpose language I know of (C/C++, Java, C#) defines this token.

There's also the issue of readability, as back-tick resemble the single-quote quite a bit. However I can definitely understand if people with an obvious primary back-tick key (US users) are tempted to use it in the alphabet, I just wanted to air the issue before Fan gets too popular and locked down.

I came to the same conclusion as Brian, because tons of languages also uses it, I know I've used it a LOT with Perl.

Im French and lived there until I was ~24 and I remember the keyboards there sometimes being an issue with a few characters (Pipe maybe, not sure) which required doing something silly like Alt-gr(right alt) then some key (or 3dgit number).... not very practical. Get an idea here: http://en.wikipedia.org/wiki/AltGr_key

On the other hand there just aren't many characters available for the grammar, as a matter of fact Fan as basically run out already and ` is used in other languages ...

One thing i remember doing back in France was using config/freeware keymapper to map say | to F12(or whatever unused key)... not perfect but it helps a lot.

It's true, PHP, Perl, Ruby and Python uses them but these are all grass-root languages with designers who did not necessarily have i18n-interoperability high on the list - hell, they are really paying for lack of Unicode now in the Ruby world.

Or let me turn it around a bit. Why is it that the current grammar is not using proper complementary tick pair, back-tick AND a fore-tick? Could it be because fore-tick is too hard to find on US keyboards? ;)

Actually I've seen the redability been an issue, in some sites i've seen the backtick look totally like a single quote(font thing?) so that can be confusing/misleading.

One thing I'm thinking, even though I'm not sold on it would be to have URI's use single quotes .... I know character already but a char is either a single letter a or unicode like \uwxyz and I don't think either could be confused with an URI.

A ' within an URI could be an issue, but do ' in URI's make any sense to start with ?

Anyway, not sure it's a good idea, but throwing it out there.

I like that, less exotic input alphabet and no accidental confusing the single-quote and back-tick. Only one problem, RFC1738 states the following:

only alphanumerics, the special characters "$-_.+!*'(),", and reserved characters used for their reserved purposes may be used unencoded within a URL.

Of course, in practice one could escape it (as browsers also do btw. before they reach the webserver) but now sure Brian is ready to go there.

BTW Guido Van Rossum(Python creator) is dutch (not sure what keyboard he uses), and I remember that the Ruby creator(Yukihiro Matsumoto) sounds like he is Japanese to me.

True, and Bjarne Stoustrup (C), Rasmus Lerdorf (PHP) and Anders Hejlsberg (C#) are Danish. Anyway, I can live with the tick thing but I do feel that it makes Fan slightly more eccentric than I would like it to be, as a potential next-gen general purpose language. That's all. :)

Are those chars a problem for anybody using non-US keyboards?

No problem at all for Japanese keyboards.

I can confirm that it's also an issue with French/Canadian keyboards though, as the back-tick button is mapped directly to "e accent grave". And this page mentions how back-tick is not available at all on Italian keybords which is confirmed by looking at the layout.

Btw. Jeff Atwood of Coding Horror / StackOverflow fame also completely forgets to mention the back-tick as a common symbol for developers, although one of the comments does mention that he succeeded to find the back-tick by accidentally holding down three keys simultaneously.

I suppose one can always do an expanding macro in NetBeans or hammer in <Alt Gr> + 0094.

What about Uri<|blah|> as a DSL plugin, as an alternative?

Damn, Italian keyboards aren't made for programming. Some people in other countries already complain about having to use Alt-Gr to get some of the often used symbols like {} or [] (personally I don't find it that hard), but holding Alt-Gr and Shift to get {} looks like finger acrobatics.

But ` not being avilable at all might be a real problem.

What about Uri<|blah|> as a DSL plugin, as an alternative?

That is an excellent idea actually, and consistent with how we do Str.

Although personally I would find this quite annoying reading code:

using Uri<|java:java.util|>
file + Uri<|dir/file.txt|>

vs

using `java:java.util`
file + `dir/file.txt`

Could one not use context sensitivity? Like what the C# compiler employs to allow LINQ expressions. For instance allow:

using java:java.util

Although that's not very easy for the parser nor particular consistent. However it seems to me that the most annoying aspect for back-tick impaired people by far, would be to handle something as frequent as import statements in Fan code - I reckon many would then employ a Copy-Paste strategy.

Btw. I've talked to a colleague today and asked what he thought of the back-tick. After explaining which one that is, his answer was "So what's next, we'll have to use ¤ and ¨ ?"

Personally, i vote to leave it as is. The ` character is common enough in enough languages and OSes (UNIX makes fairly heavy use of it, e.g.). The Uri<||> syntax just looks unwieldy for something that's supposed to be a first class member of the language.

What about something like @"http://blahblahblah"?

Could one not use context sensitivity?

In that particular case yes, but not in the overall grammar. So from a tooling perspective, much easy to keep literals as tokens versus breaking them up into multiple tokens.

Btw. I've talked to a colleague today and asked what he thought of the back-tick. After explaining which one that is, his answer was "So what's next, we'll have to use ¤

Just to note - the huge difference between other chars and the back-tick is that the back-tick is a proper 7-bit ASCII char.

Personally, i vote to leave it as is. The ` character is common enough in enough languages and OSes (UNIX makes fairly heavy use of it, e.g.).

I agree. Not to discount our disadvantaged friends across the pond, but it seems that many languages use this symbol okay. Anything else would probably be harder to read.

This doesn't prevent us from adding another syntax in the future.

And another aspect of this problem, is that seems like it should be easily solved with by the editor. For example something like the NetBeans just make a nice macro for Uri literals?

I don't mean to trivialize your issue casper, but after hearing the feedback here, I believe we should stick with using the back-tick. Uri's are so fundamental in Fan, I think its worth having a dedicated delimiter char.

Fair enough, I understand that status quo is highly advantageous. I assume tooling support would be able to auto-import using statements just like it works with Java import statements so that should take care of the most common usage pattern. Thanks for hearing me out. :)

Interestingly I've started working on import completion/fix in my netbeans plugin last night ... but it will be maybe ~2 weeks before I have to time to finish it.

Not sure i see how that helps with the backticks though.

Thanks for hearing me out. :)

No problem - thanks for raising the issue.

Not sure i see how that helps with the backticks though.

Because I am going to be changing Java FFI to use URI literals:

using `java:some.package`

Ho yeah right, forgot about that already :) When are you gonna change this in HG ? and ~how long before it's released.

Don't need hard dates, just a general idea to help me decide if I should bother with the "old" syntax or just go straight for the new one.

Ho yeah right, forgot about that already :) When are you gonna change this in HG ? and ~how long before it's released.

Since we need it for some SkyFoundry work, probably either this week or next.

I quite like the backtick, but I have a nice UK keyboard. I don't like URI<||>. I wouldn't mind single quotes, or an @"htttp://..." notation, although I'm aware that the single quotes version conflicts with fandoc.

Just to clarify, I was suggesting Uri<|something|> syntax as an alternative syntax, not a replacement. The same way you can code either "foo" or Str<|foo|> and it means the same thing.

Out of curiosity, does Fan validate URIs at compile time?

Out of curiosity, does Fan validate URIs at compile time?

The URI literals used in Fan are what I term "standard form", and there isn't an incorrect syntax - anything goes with ambiguity in separators handled via backslash. This differs from "encoded" form which is what you would use over HTTP.

See the fandocs for sys::Uri for how it all works.