Rb_eval_string_protect with correct error line number

ruby-talk
1

Rubies:

rb_eval_string_protect() calls eval(), so at error time $ERROR_POSITION ($@)
gives me a useless number. I want the line number (and preferably the column
number) of the offending part of the evaluated string.

Raw Ruby does that like this:

    stwing = """
        puts 'a'
        puts undeclaredVariable
        puts 'c'
    """

    begin
    eval(stwing, nil, "yo", 0)
    rescue
      p $@
    end

That outputs ["yo:2"]

How can I get the correct error line number, using the equivalent of
eval()'s last two optional parameters, without converting my string to a
Ruby string and then calling "eval" on it?

···

--
  Phlip
  http://industrialxp.org/community/bin/view/Main/TestFirstUserInterfaces

2

Hi,

···

In message "Re: rb_eval_string_protect with correct error line number" on Fri, 24 Sep 2004 10:54:32 +0900, "Phlip" <phlip_cpp@yahoo.com> writes:

How can I get the correct error line number, using the equivalent of
eval()'s last two optional parameters, without converting my string to a
Ruby string and then calling "eval" on it?

There's no way with the current implementation. Use double eval for
the purpose.

              matz.

3

Phlip wrote:

This code embeds Ruby, executes a string found in 'source', and fetches any
error message:

    rb_gv_set("$errorLineNumber", Qnil);
...

    VALUE value = rb_str_new2( source );
    rb_gv_set("$evalMe", value);

    rb_eval_string_protect(
            "begin\n"
            " $errorLineNumber = nil\n"
            " eval($evalMe, nil, 'eval', 1)\n"
            "rescue\n"
            " $errorLineNumber = $@[0].slice(5..100).to_i()\n"
            " p $!\n"
            " p $@\n"
            " raise\n"
            "end\n"
            , &state);

    if (state) return state;
...
    VALUE errorLineNumber = rb_gv_get("$errorLineNumber");

    if (errorLineNumber != Qnil) // else it's in a library or something
hard
        {
        int lineNo = NUM2INT(errorLineNumber);

Next problem: 'source' may contain free-form Ruby code, entered by a user.
If the user creates a variable 'foo', executes the code, erases 'foo', and
executes again, 'foo' still exists on the heap.

I want an error message if 'foo' no longer exists, from the user's point of
view. How do I reset Ruby back to the state before the eval() call? I tried
passing eval() a Proc.new{}, and I tried terminal functions like
ruby_prog_init(), ruby_finalize(), and ruby_cleanup(). They seem to destroy
too much.

So, how to clean up a local stack?

···

--
  Phlip
  http://industrialxp.org/community/bin/view/Main/TestFirstUserInterfaces

4

Hi,

So, how to clean up a local stack?

How about wrapping user code in the non-iterating block?
E.g.

  loop {
    ..user code..
    break
  }

The local variables defined within a block will disappear, when the
block execution terminated.

              matz.

···

In message "Re: How to clean out my stack" on Sat, 25 Sep 2004 03:10:03 +0900, "Phlip" <phlip_cpp@yahoo.com> writes:

5

Yukihiro Matsumoto wrote:

>So, how to clean up a local stack?

How about wrapping user code in the non-iterating block?
E.g.

  loop {
    ..user code..
    break
  }

That did it - thanks!

BTW major props for this bitchen language, holmes.

I'm writing a front-end for this:

    http://flea.sourceforge.net/

Those shapes come from command sets like this:

     b.tube.shorter(0.8).left.link(b)

That tells the 'b' chromosome (axiom, turtle, whatever) to draw a tube, then
get shorter, bend to the left, and recurse into itself. The result (rendered
in POVray) is a spiral.

    http://flea.sourceforge.net/link_example.png

Traditionally, you write a file called something.flea, and run a command
line script to render it. I'm writing an editor to put the Flea commands
into a panel on the left, and an OpenGL output on the right. So this
requires me to embed Ruby inside this editor. I could have also written the
editor in Ruby, but I chose C++ because I'm a masochist.

···

In message "Re: How to clean out my stack" > on Sat, 25 Sep 2004 03:10:03 +0900, "Phlip" <phlip_cpp@yahoo.com> writes:

--
  Phlip
  http://industrialxp.org/community/bin/view/Main/TestFirstUserInterfaces