Re: Scheme question on strict substitution

[stk]

Hello,

As regards the issue of context-free equivalence --

> > I can say
> >
> >    foo = \markup { \bold "Zanzibar" }
> >    bar = \markup { "Stand on" \foo }
> >
> > but I can't say
> >
> >    bar = \markup { "Stand on" \markup { \bold "Zanzibar" } }
> >
> > [so]
> >
> >    \foo   is not equivalent to   \markup { \bold "Zanzibar" }
> >
> > despite the "=" sign in the macro definition

> The reason for this is that foo is not a macro; it's a variable.
> Similarly, in C you can write
> int i = 5+9;
> foo(&i);
> which is _not_ the same as the (invalid) expression
> foo(&(5+9));
> (this example is not exactly analogous to what you're trying, but it
> illustrates that variables and macros are different things).

No, it illustrates that variables and literals are different things.
In C, "&" means "contents of" an address, and  i  is an address, but
(5+9) is not, so you can't write "&" in front of (5+9).  But in LilyPond,
"\" in "\foo" means eval(foo), where foo is a variable, but "\" in
"\markup" is not an eval operator, or any other kind of operator, it is
just a character used lexically to indicate a keyword.  There will
inevitably arise cases where users are confused about acceptable syntax
because of the various different uses of "\" (consider also 2\cm).

> The special thing about 'markup' is that it isn't a function; it's rather a
> special token that marks that the following code should be interpreted as
> markups.

If that's all there were to it, then

      bar = \markup { "Stand on" \markup { \bold "Zanzibar" } }

would work.  The special thing about \markup is that it is analyzed
strictly lexically:  it can have only certain lexical items in it, like
\bold, \raise, \column,...   Given that as a starting point, it must have
been very difficult to write the additional code to allow it to contain
also constructs like \foo (where foo is a defined variable).  If it had
been analyzed dynamically, then *both* the above expression and

     bar = \markup { "Stand on" \foo }

would work, since   \foo   and   \markup { \bold "Zanzibar" }   both
produce exactly the same stencil as a result.  Maybe I'm griping too
much, maybe that's what you meant by saying that \markup isn't a function.

I must sound ungrateful, nitpicking about syntactical inconsistencies.
To be more reasonable, what I have learned by using *any* typesetting
program is that music layout is wildly difficult.  It is almost
unbelievable that Han-Wen, Jan, and others (I'm sorry I don't know all the
names) managed to create a working system to produce such beautiful
results.  Certain parts of LP syntax are hard to be sure of when you're
writing a   .ly   file, but the gorgeous printed output never ceases
to amaze me.

-- Tom

*******************************************************

On Fri, 1 Dec 2006, Erik Sandberg wrote:

> On Thursday 30 November 2006 21:32, address@hidden wrote:
> > Hello,
> >
> > > . . . one problem is that this [giving the syntax of each keyword] would
> > > still just tell a small part of the full syntax.
> >
> > I think it would give a big part of the full syntax, even if not the whole
> > picture.  It would enable a user to know
> >    (1) what type of "arguments" have to follow the keyword;
> >    (2) whether (s)he has accidentally omitted one of the required
> >        arguments (when debugging);
> >    (3) whether (s)he should write a macro definition or a function
> >        definition to specify a "\whatever" construct involving the
> >        keyword;
> >    (4) what the types should be in the event that (s)he winds up writing a
> >        function definition.
> > "Full" or not, that's a lot of useful information.  And I'm talking about
> > just a list of the syntax for each keyword, *not* the semantics.  A syntax
> > specification can be short but still tremendously useful.  Specification
> > of the semantics takes a lot of blah-blah, and nobody would have the time
> > to write up all the semantics.
> >
> > > Another problem is that the
> > > rule I gave really was too simple. For example, you cannot say
> > > ^"some text" in a .ly file
> > > without having a note in front of it, still you can define a macro
> > > mytext = ^"some text"
> > > and then use it as c \mytext
> >
> > That's a general problem in LilyPond, knowing what kind of expression
> > constitutes a valid right-hand side of a macro definition.  (I'm amazed
> > that the above definition works.)
> >
> > A different but related problem concerns something I don't know the right
> > term for, something like "context-free equivalence" maybe:  for example,
> > I can say
> >
> >    foo = \markup { \bold "Zanzibar" }
> >    bar = \markup { "Stand on" \foo }
> >
> > but I can't say
> >
> >    bar = \markup { "Stand on" \markup { \bold "Zanzibar" } }
> >
> > You will object that I would have no *reason* to say that anyway,
> > but it nevertheless constitutes a toy example of non-transferability:
> >
> >    \foo   is not equivalent to   \markup { \bold "Zanzibar" }
> >
> > despite the "=" sign in the macro definition, and that's surprising.
>
> The reason for this is that foo is not a macro; it's a variable. Similarly, in
> C you can write
> int i = 5+9;
> foo(&i);
> which is _not_ the same as the (invalid) expression
> foo(&(5+9));
> (this example is not exactly analogous to what you're trying, but it
> illustrates that variables and macros are different things).
>
> The special thing about 'markup' is that it isn't a function; it's rather a
> special token that marks that the following code should be interpreted as
> markups.
>
> --
> Erik