Re: [Python-markdown-discuss] Bug 1763338 and Max Recursion

[Waylan L. <wa...@gm...>]

Nevermind. This is now fixed in r57. Nested listes were giving me a
few headaches that seemed obvious with recursion, but not with a loop.
I worked them out though.

On Nov 16, 2007 11:20 PM, Waylan Limberg <wa...@gm...> wrote:
> Some time ago I looked at bug 176338 [1] and lowered its priority
> significantly as the sample input was a far cry from markdown syntax.
> I recognized the need for a more graceful way to deal with the issue,
> but given the formatting of the text, it didn't seem all that
> important.
>
> Thing is, now I realize that the sample text attached to that ticket
> was skewing the issue. The bad syntax was just escalating the problem.
> The fact is, the current way `_process_section` works (it recursively
> calls itself on each block of remaining text), we can easily reach max
> recursion with perfectly good syntax.
>
> Therefore, I created a document the following way:
>
>     import sys
>     t = ""
>     for i in range(sys.getrecursionlimit()):
>         t += "Paragraph %d.\n\n" % i
>
> That's 1000 paragraphs one line each - as simple of markdown syntax as
> one could have. Even so:
>
>     html = markdown.markdown(t)
>     ...
>     RuntimeError: maximum recursion depth exceeded
>
> I played with a few different numbers and 985 was the magic number.
> The maximum paragraphs a markdown document can have is 984. That's
> assuming no lists, blockquotes or anything else that necessarily goes
> deeper into recursion.
>
> While replacing recursion with a loop is probably easier said than
> done, do we care? In other words, are we ok limiting the possible
> length of documents we can process - with that limit becoming shorter
> the more complex the syntax used?
>
> If so, I'll just catch the exception and output a friendlier message.
>
> If not, we may have some work to do - although I'd suggest not holding
> up 1.7 for it.
>
> [1]: http://sourceforge.net/tracker/index.php?func=detail&aid=1763338&group_id=153041&atid=790198
>
> --
> ----
> Waylan Limberg
> wa...@gm...
>



-- 
----
Waylan Limberg
wa...@gm...