silgraphite-fonts

[Silgraphite-fonts] GDL question: Matching and replacing two subsequent identical characters?

[Carsten B. <ca...@gm...>]

Hi,

I've got another question regarding GDL and substitution: How do I match 
and replace two subsequent identical characters? What I want to do is 
basically this:

   Consonant Same_Consonant > Consonant Gemination_Diacritic

Since according to the documentation it's not possible to use 
referencing with @ or $ on the LHS, I'm a little stumped. Is this maybe 
solvable with some flag, or a recursive rule?

Regards
Carsten

-- 
Currently developing Tagāti Book G: http://bit.ly/Ml5xb8

Re: [Silgraphite-fonts] GDL question: Matching and replacing two subsequent identical characters?

[Sharon C. <sha...@si...>]

There's not really a good way built into the language--normally you'd 
have to list all the pairs using separate rules.

But here's another idea:

table(glyph)

g_b {type = 1};
g_c {type = 2};
g_d {type = 3};
g_f {type = 4};
etc.

Consonant = (g_b, g_c, g_d, g_f, ...);

endtable;

table(sub)

Consonant  Consonant  >  Consonant  Gemination_Diacritic  /  _ {type 
== @2.type}  _ ;

endtable;

Actually making a set of separate rules is probably slightly more 
efficient--using a finite state machine to match rules is more 
efficient than testing the glyph attribute. But maybe the code isn't 
as neat. Although either way you have to make a long list.

On 8/5/2012 7:52 AM, Carsten Becker wrote:
> Hi,
>
> I've got another question regarding GDL and substitution: How do I match
> and replace two subsequent identical characters? What I want to do is
> basically this:
>
>     Consonant Same_Consonant > Consonant Gemination_Diacritic
>
> Since according to the documentation it's not possible to use
> referencing with @ or $ on the LHS, I'm a little stumped. Is this maybe
> solvable with some flag, or a recursive rule?
>
> Regards
> Carsten
>

Re: [Silgraphite-fonts] GDL question: Matching and replacing two subsequent identical characters?

[Shriramana S. <sa...@gm...>]

Hi Sharon (or Martin and others),

On Sun, Aug 5, 2012 at 6:22 PM, Carsten Becker <ca...@gm...> wrote:
> Since according to the documentation it's not possible to use
> referencing with @ or $ on the LHS

Is there a prohibitive reason by which it is not possible to use
referencing in the context? I seem to have encountered a situation
similar to Carsten's some time back...

@Carsten: I think it would be $ and not @ too which references the
whole glyph and not just its index.

-- 
Shriramana Sharma

Re: [Silgraphite-fonts] GDL question: Matching and replacing two subsequent identical characters?

[Martin H. <mar...@si...>]

Dear Shriramana,

> Is there a prohibitive reason by which it is not possible to use
> referencing in the context? I seem to have encountered a situation
> similar to Carsten's some time back...

There is nothing to stop us making a decision to support this. But the cost would be considerable and outweighs the benefit. The regular expression engine is deterministic and adding back references would cause a radical change involving rewriting nearly everything in order to turn the regular expression non deterministic.

One option might be to add a slot attribute .glyphid or somesuch that could be tested:

cons > geminate / cons _ {glyphid == @1.glyphid};

But that would take a while to appear (since it would have to be added to the compiler and the engine and then trickle down via releases) and be pretty slow, so perhaps you would be better off with doing something like:

#define GEM(x, y) x > y / x _

GEM(g_a, g_a_gem);
GEM(g_b, g_b_gem);

etc.

That will run much faster than any constraint test would.

Yours,
Martin

Re: [Silgraphite-fonts] GDL question: Matching and replacing two subsequent identical characters?

[Shriramana S. <sa...@gm...>]

Hi Martin -- thanks for your reply.

On Mon, Aug 6, 2012 at 9:09 AM, Martin Hosken <mar...@si...> wrote:
> One option might be to add a slot attribute .glyphid or somesuch that could be tested:
> cons > geminate / cons _ {glyphid == @1.glyphid};

Another option might be to allow some kind of foreach construct in the
language. But perhaps that would be somewhat deviating from the
existing language model. And without LHS referencing, it still would
not be useful for much more than geminates. (Neither would glyphid
slot attributes.) Although, honestly I cannot imagine what real case
other than geminates would need to be supported.

In my Tamil Brahmi GDL, I had to do 18 series of:

KA + clsVowelSigns -> KA-series
NGA + clsVowelSigns -> NGA-series

and so on. (But that was because the glyph developer had made
pre-composed glyphs which I realized later are not necessary.)

But if GDL had for and treated classes like arrays, we could always
use a two dimensional array:

for x = 1 to 18: for y = 1 to 12:
 cons[x] vs[y] > syllable[x][y]

;-)

> #define GEM(x, y) x > y / x _
> GEM(g_a, g_a_gem);
> GEM(g_b, g_b_gem);
> etc.

Well if it came to typing everything out, this is only a little less
cumbersome than the full version. I'm sure right now we can use Python
(or some such scripting language) to output the requisite statements
using *its* for loop.

-- 
Shriramana Sharma

Re: [Silgraphite-fonts] GDL question: Matching and replacing two subsequent identical characters?

[Carsten B. <ca...@gm...>]

Am Mo 06 Aug 2012 09:03:54 CEST schrieb Shriramana Sharma:

> Well if it came to typing everything out, this is only a little less
> cumbersome than the full version.

I typed it all out in the end. And while that certainly looks more 
cumbersome than a few lines of determining character properties on the 
fly, it's still just 30-odd lines like the one below, so that's entirely 
manageable.

     U+0070 AnyVirama U+0070 > @1:(1 2) _ U+0313; // PA (key: p)

Carsten

-- 
Currently developing Tagāti Book G: http://bit.ly/Ml5xb8