[Libguestfs] [PATCH 2/3] v2v: add Var_expander

[Pino Toscano]

On Monday, 25 March 2019 15:55:46 CET Richard W.M. Jones wrote:
> On Mon, Feb 25, 2019 at 05:22:51PM +0100, Pino Toscano wrote:
> [...]
> 
> After being burned a few times with custom parsing (hello, guestfish)
> I'm not a big fan.

I can perfectly understand that, however ...

> Is there not an existing C or OCaml library/facility we could use
> here?  It's a shame we can't use Perl Template Toolkit because it
> would be ideal here.

... sadly I did not find anything simple convering the use case that
this Var_expander module covers.  The closest thing I found was the
usage of ${..}/$(..) variables in dune (the OCaml build system), with
the following differences:
- manually tokenizes the string
- allows anything as variable name, splitting it in two if a ':' is
  found
- ${..}/$(..) instead of %{..}

> There are all kinds of questions that aren't answered such as: Should
> variables be replaced recursively?
> [...]
> Should we allow loops or similar constructs?

I don't think they are needed.

> How do you escape %{..} if you don't want it to be replaced?

Good question... dune does not seem to allow that; maybe we can allow
a trailing '$'/'\' as escape character.

> Existing template systems solve these kinds of problems already.

Yes, however they are much more complex, and with a number of features
like conditionals, filters, loops, etc.  Considering this is needed so
far for the file paths of disks in -o json, IMHO plain variables are
enough.

> > +let var_re = PCRE.compile "%{([^}]+)}"
> 
> Are we planning to allow a completely free choice for variable names,
> or could we limit this regexp to only matching ASCII alphanumeric +
> underscore?

This regex allows anything as variables for two reasons:

1) we do not miss any variable-like pattern (so we do not silently
   skip any now, while handling it in the future in case we accept
   more characters)

> > +let check_variable var =
> > +  String.iter (
> > +    function
> > +    | '0'..'9'
> > +    | 'a'..'z'
> > +    | 'A'..'Z'
> > +    | '_'
> > +    | '-' -> ()
> > +    | _ -> raise (Invalid_variable var)
> > +  ) var
> 
> ... and then this function would presumably go away.

2) we can check that a variable has only allowed characters, and report
   that to the user

-- 
Pino Toscano
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: This is a digitally signed message part.
URL: <http://listman.redhat.com/archives/libguestfs/attachments/20190329/0124d0b1/attachment.sig>