1. it requires a "special" undocumented view in Search.pm to work
2. the topics matched by that initial search are different to the topics matched by the actual process that does the grep, so there is a real risk of missing topics.
3. because it uses SEARCH, it cannot ignore NOSEARCHALL
4. it's way over-complicated. It took me ages to work out exactly how it worked.

I have cured all these problems - and more - as follows:

1. recoded the initial search so it is done from code instead of a variable. Sounds more expensive, but it actually saves quite a lot of code because the renameview is no longer needed, so
2. removed the renameview from Search
3. generalised the replacement code into a forEachLine parser method that iterates over the lines in a topic, tracking pre and verbatim etc.
4. shared the "wikiword recognition" code with the actual rendering code, so it's harder to get confused over what is a wikiword and what isn't
5. added a unit test for the rename process.

The only change from an end user perspective is a change to the way the search is presented; it has become a lot more comapact.

For a demo, see ntwiki, or svn up your checkout area.

-- CrawfordCurrie - 18 Apr 2005

Looks better. Did you remove the last topic's author on purpose?

-- ArthurClemens - 18 Apr 2005

Yeeeees, sort of. Basically I radically simplified the search.

My feeling is that if you wanted that sort of detail, you would do a proper search, and it's just noise in a rename view.

But the main difference is that now the rename code has a consistent view of "what is a wikiword" as the rendering code.

-- CrawfordCurrie - 19 Apr 2005