Re: Termination of variable definitions

[David Kastrup]

David Sumbler <address@hidden> writes:

> On Tue, 2016-04-19 at 20:51 +0200, David Kastrup wrote:
>> David Sumbler <address@hidden> writes:
>> 
>> > In starting to experiment with selective compilation using includes
>> > and/or the $(if condition action) structure provided by David Kastrup, I
>> > found that there is something very, very basic that I have never really
>> > understood.
>> >
>> > So at the risk of embarrassing myself, I should be grateful if somebody
>> > would explain the following: how does Lilypond recognize the end of a
>> > variable definition?
>> 
>> One expression.
>> 
>> > For instance, an example from the documentation shows: 
>> >
>> > violin = \new Staff {
>> >   \relative {
>> >     a'4 b c b
>> >   }
>> > }
>> 
>> "cello" can be no part of a single music expression, so it becomes a new
>> statement.  Actually, at this point there are very few possible
>> continuations.  But there is one:
>> 
>> \addlyrics { oh right this one }
>> 
>> > cello = \new Staff {
>> >   \relative {
>> >     \clef "bass"
>> >     e2 d
>> >   }
>> > }
>> >
>> > What I don't understand is why the definition of 'violin' extends from
>> > 'new' to the right brace on line 5.  Why does the definition not extend,
>> > for instance, to the brace on line 12?
>> 
>> What would "cello" be?
>> 
>> > Alternatively, why does 'violin' not equate simply to '\new' or to
>> > '\new Staff' or '\new Staff {' ?
>> 
>> All those are not a complete music expression.
>
> So Lilypond actually checks the syntax and structure of a variable at
> the time it is defined.

It does not "check" it as much as read it in and evaluate it completely.

If some expressions occur at top level, it does a default action.
Scores a collected into the default book which is typeset at the end of
run when non-empty.  Books are typeset, bookparts are evaluated and
stored in the default book and so on.

But LilyPond is an interpreter, not a compiler.  Expressions get parsed
when they are encountered.

> I assumed that only the validity of the name was checked, and that the
> rest was simply stored as a string of characters.  This is implied in
> the docs,

Where?  It would be completely wrong.

> although it may well be that elsewhere they do state that the string
> must represent a complete music expression.

There is no string here.  Quite literally not.  The lexer turns the
input into tokens, and the parser collects the tokens into productions.
Only in a minority of cases is an actual Scheme string produced from
them: most of the time they are just characters in the input buffer.

> Also from the documentation we have:
>
> width = 4.5\cm
> name = "Wendy"
> aFivePaper = \paper { paperheight = 21.0 \cm }
>
> None of these, as I (perhaps mis)understand it, is a music expression.

I wrote "one expression", not "one music expression".

> I can see that the definition of aFivePaper is perhaps terminated by
> the '}' following the earlier '{'.  And that the definition of name is
> perhaps terminated by a double quote, since it commences with one.  In
> the case of width, it seems even less clear - perhaps it is that, in
> the absence of an opening delineator such as ( { " etc., it regards
> the first white space as a terminator.  But I am guessing.
>
> Am I alone in finding the functioning of variable definition slightly
> less than obvious?

It's variable = expression and LilyPond knows a variety of different
expressions.  The expression is read in, and when it is finished, it is
assigned to the variable with name "variable" (in this example).

-- 
David Kastrup