Re: Open Source
don cole wrote:
>
> Now that Euphoria open source. (The answer to everything.) How come we now
> can't have a Goto?
>
> Don Cole
There has been sustained opposition to it in the past, on the premise that it
promotes unmaintainable code.
My take on this s that either:
1/ If there were more control flow structures, like exif (exit current if
statement), next (evaluate next loop conditional clause), retry (reatsrt at top
of loop, skip conditional clause), and if all those, as well as exit, could have
an optional parameter, either a relative integer or a label string (to exit
nested blocks), then goto wouldn't be needed. That would be my preference.
2/ Some will object the extra keywords. Another possibility is to have a managed
goto: goto statements will need a label, so that come_from() will return which
label was jumped from last, and come_back() would return to statement following
last taken goto. This scheme is exposed in gretar details in the OpenEu
specifications at http://oedoc.free.fr .
3/ Yet another: define a new "with RAD" directive, off by default, that would
turn on constructs deemed of use at development stage only, not release stage.
goto and assert() would be candidates for such constructs.
4/ And after all, I don't care. If someone can use goto effectively, it should
be there; if someone uses it to release poor code, that code will be hardly
maintained and will go out of sight in a darwinian way.
Note that, if goto is allowed, then there must be some way to deal with a goto
made inside a for loop, because of the special scoping rules for loop indexes.
goto-ing into a routine should not be allowed. goto-ing into another file's top
level code may be useful at times, but has issues of its own. See the goto_far
statement in OpenEu specs, and also discussions on the OpenEu list on Topica.
CChris
|
Not Categorized, Please Help
|
|
