Re: Fix #887: Use ly:string-percent-encode for textedit URIs. (issue193077)

[pnorcks]

Reviewers: lemzwerg, dak,

Message:
On 2010/01/25 09:28:36, dak wrote:
> http://codereview.appspot.com/193077/diff/1001/12
> File lily/general-scheme.cc (right):

> http://codereview.appspot.com/193077/diff/1001/12#newcode230
> lily/general-scheme.cc:230: return ((c >= 0x2D && c <= 0x2F) //
hyphen,
> full-stop, and forward-slash
> On 2010/01/25 06:38:18, lemzwerg wrote:
> > Wouldn't it be faster to use an array of `0' and `1', indexed by the
character
> > code, instead of the many comparison operations?  E.g.
> >
> >   escape_character[128] = { 0, 0, 0, ..., 1, 1, 0, ... }
> >

> Why not let the compiler bother about this?  Note that with current
computer
> architectures, a random memory access can easily take 10 times or more
than a
> comparison.

> But I think that gcc is a compilation requirement anyway,
> so it would just do to write:

> switch (c) {
> case '-':
> case '.':
> case '/':
> case '0'...'9':
> case 'A'...'Z':
> case '_':
> case 'a'...'z':
>    return 1;
> }
> return 0;

> This saves having to write any comment since the code is clear enough
on its
> own, and the compiler has all the info it needs to generate more
concise code
> than a human could.  In particular since the optimal code may well
depend on the
> target architecture.

Thank you Werner and David for your suggestions.

I decided to go with David's suggestion to use the GCC-specific syntax.
It looks cleaner, and it reduces the need for code comments.

I'll let this patch sit here a while longer until I have a chance to do
more extensive testing.

Thanks,
Patrick

Description:
Fix #887: Use ly:string-percent-encode for textedit URIs.

* Add an overloaded instance of String_convert::bin2hex optimized for
  converting single bytes to hex.

* Add a new callback, ly:string-percent-encode, to be used for percent
  escaping paths in textedit URIs.  This does the following:

  - Leave unreserved characters in textedit URIs unescaped.  This
    includes 0-9, A-Z, a-z, and three punctuation characters (hyphen,
    underscore, and full-stop).

  - Leave the forward slash (/) unescaped, since it is used as a path
    delimiter.

  - Escape all other characters.  Don't check for a null byte, since
    those likely won't sneak into a full pathname.

* Use the callback function in the PS backend.

Please review this at http://codereview.appspot.com/193077/show

Affected files:
  M flower/include/string-convert.hh
  M flower/string-convert.cc
  M lily/general-scheme.cc
  M scm/output-ps.scm