Re: [RFC] [PATCH] Automatically quote the arguments to an eval macro

[Aaron Ecay <aaronecay@gmail.com>]

Hi Nicolas,

Thanks for your feedback.

2015ko azaroak 5an, Nicolas Goaziou-ek idatzi zuen:

[...]

>> This means:
>> 1. Users need to remember to put quotes around $n all the time
>> 2. It’s impossible to pass arguments with a " character to a macro
> 
> This can be fixed by removing the `literal' optional argument from
> `replace-regexp-in-string' call in `org-macro-expand'.

I’ve actually managed to make it work with the code as-is, by escaping
the quote with a single backslash.  I must have been confused about
single- vs. double-escaping.

The inconsistency that remains is that quotes need to be backslash-escaped
for eval macros, but not regular macros.

Here’s a test case (works with current master):

,----
| #+macro: foo $1
| 
| {{{foo(xyz\"abc)}}}
| 
| #+macro: bar (eval "$1")
| 
| {{{bar(xyz\"abc)}}}
`----

Ascii export yields:

,----
| Foo is: xyz\"abc
| 
| Bar is: xyz"abc
`----

> 
>> The attached patch changes the behavior of eval macro arguments so that
>> $1 etc. expand to the argument with quotation marks.  That is, the
>> following is now the correct way to write a macro (note lack of "s):
>> 
>> #+macro: identity (eval $1)
> 
> I'm not totally opposed to it but it introduces a limitation: all
> arguments must be strings. This is not strictly required, actually. As
> a consequence, I slightly prefer fixing the current situation instead.
> 
> Is there any strong reason to force string?

I’d say the reason is that (AIUI) macros are for string manipulation
during export.  I’m hard-pressed to come up with a good usage of a macro
that treats its arguments as non-strings.  But maybe that’s just a
function of my limited creativity. :)

> 
>> * lisp/org-macro.el (org-macro-expand): Automatically quote the
>> arguments to an eval macro.
> 
> You would also need to update `org-macro-initialize-templates' and
> `org-export-as'.
> 
>> +			(cond
>> +			 (evalp (format "%S" arg-val))
>> +			 (arg-val arg-val)
> 
> Nitpick: this is equivalent to (arg-val)

I didn’t know this usage of cond....

Thanks again,

-- 
Aaron Ecay