ExternalMethods.

is still a hack but is much cleaner (ie, in it's own module).

are intern'ed, untextindex.__getitem__ modified to look up word in lexicon

and vocab methods, and put resultlist in it's own module

wrong context.

acquired objects.  getRolesInContext did not used the correct context,
as defined by the inner-most object wrapping.

Notepad? :)

data_record_id_ attribute, brains now also have a data_record_score_
attribute.  Currently, the score is 1 for objects returned from field
indexes, and search term frequency for text indexes.

the cache.  Modified UI to include CPU time and enable/disable subtransactions

the cache

Also changed completion code to check registered status and
avoid completion logic if not registered, which would be the case if
_begin failed.

added to try to reduce computational overhead only works correctly for
HTTP and broke FTP and other stateful protocols. I took the caching back
out - maybe we can try to C-ify the method at some point as a better
alternative.

Fixed a find support bug which caused find permissions not to get correctly registered. Also fixed a find frame URL bug that became apparent after fixing the first bug.

Thanks Brian Hooper.

Common Long Format.

reload extenal method files every time.

The problem was that the entire os.stat() record was being compared,
and that includes the access time, which of course changes every time
the method is loaded - self fulfilling prophecy, and all.

Instead, we now just stash and compare the modification time
(ST_MTIME) field, and get rid of the extraneous (expensive) reloads.

ZTables)

that requires climbing the entire aq chain on every call (and it gets
called a *lot*). A little investigation revealed that it is called
over and over for the same objects during a request for complex pages,
so a cache was added that can cut the work down to about 10% of its
current level.