Re: scheme-function to provide value for \include

[David Kastrup]

Urs Liska <address@hidden> writes:

> Am 28.06.2016 um 11:05 schrieb David Kastrup:
>> Urs Liska <address@hidden> writes:
>>
>>> Hi all,
>>>
>>> I am surprised that \include refuses to accept a string that isn't
>>> passed literally but through a scheme-function:
>>>
>>> \version "2.19.43" givePath = #(define-scheme-function ()()   
>>> "some/path/that/will/probably/not/be/found.ily") #(display (givePath))
>>> \include \givePath
>> \include is not a scheme function.  It switches input at a very low
>> level.  \include is not even a reserved word and it never gets turned
>> into a token.  It is a pattern of bytes causing an immediate reaction in
>> the lexer.  That it accepts a #xxx or $xxx expression is hard-coded
>> right in the lexer: it does not pass through the normal parser
>> mechanisms for these constructs.
>>
>> Basically you are expecting something akin to the #include of the C
>> preprocessor to accept calls of functions defined in C for specifying
>> the file to include.
>>
>
> Thank you for the explanation.
> Interestingly Harm provided a solution presumably working,

"working" is not the same as "doing something".  To wit:

    givePath =
    #(define-scheme-function ()()
      "some/path/that/will/probably/not/be/found.ily")

    \include #(givePath)

givePath is not executed by the parser at all but merely called from
Scheme, which makes the above a complicated way of writing

#(define (givePath) "some/path/that/will/probably/not/be/found.ily")

and using Scheme for calculating an expression to \include .  Nothing is
interpreted by LilyPond's parser here with regard to the \include .

-- 
David Kastrup