[prev in list] [next in list] [prev in thread] [next in thread] 

List:       zope-coders
Subject:    Re: [Zope-Coders] Patch for KeywordIndexes
From:       Casey Duncan <casey () zope ! com>
Date:       2003-05-31 3:00:28
[Download RAW message or body]

All of the type tests in here are kludges. 

How about something like this:

from types import StringType, UnicodeType

...

if (isinstance(newKeywords, StringType) 
    or isinstance(newKeywords, UnicodeType)): #Python 2.1 compat isinstance
    return (newKeywords,)
else:
    try:
        newKeywords = tuple(newKeywords)
    except TypeError:
        # Not a sequence
        return (newKeywords,)
    else:
        ... uniqueify ...

or as a variation of your code:

if (isinstance(newKeywords, StringType) 
    or isinstance(newKeywords, UnicodeType)): #Python 2.1 compat isinstance
    return (newKeywords,)
else:
    unique = {}
    try:
        for k in newKeywords:
            unique[k] = None
    except TypeError:
        # Not a sequence
        return (newKeywords,)
    else:
        return unique.keys()

I know we're stuck with this dwim for bw compatibility, but I hope that we can 
consent to simply require that the source be a sequence in future "keyword" 
index implementations.

-Casey

On Friday 30 May 2003 07:13 pm, Sidnei da Silva wrote:
> On Fri, May 30, 2003 at 05:34:30PM -0400, Tres Seaver wrote:
> | On Fri, 2003-05-30 at 17:21, Sidnei da Silva wrote:
> | > If no one object (or if someone provides a better way to do this) I
> | > would like to apply the following patch for KeywordIndexes on
> | > 2_6-branch and HEAD. Its just completely annoying that this worked
> | > before, and now its so stupidly broken :(
> | 
> | I don't think your patch works correctly:  the test for "string-like" is
> | not the same as a test for iterability.  As a check, try indexing an
> | object with a "string" property (instead of "lines" or "tokens").  I'll
> | bet you end up indexing each letter of the string as a "keyword".
> 
> You're right. It needs to test for string-like *and* for
> iterability. Here's the fixed version.
> 
> -- 
> Sidnei da Silva (dreamcatcher) <sidnei@x3ng.com.br>
> X3ng Web Technology <http://www.x3ng.com.br>
> GNU/Linux user 257852
> Debian GNU/Linux 3.0 (Sid) 2.4.20-powerpc ppc
> 
> Real Programmers don't eat quiche.  They eat Twinkies and Szechwan food.
> 



[prev in list] [next in list] [prev in thread] [next in thread]