Hello Michael, MK> Then you end up using regex again, I don't know, but the phrase "more MK> reliable" than just switching out array values in the instructions list kinda MK> doesn't sound right to me here ;-). Yeah, I know. Regex's are evil. Wait, doesn't lexer use them to parse the pages? ;) Though, I do agree that "more reliable" was quite unfortunate choice of words :) MK> Can you give an example of what you're doing with the instructions? Do you add MK> extra plugin instructions (that should be no problem). Or do you something MK> different? In Columns plugin I move around section_[open|close|edit] instructions to improve interaction between section <div> tags and the columns <table>. I guess it should be benign as this instructions have no text attached to them. In RefNotes I run lexer recursively on a piece of wiki text and stuff resulting instructions into a nest. The nested block triggers the plugin to capture rendered output to be used elsewhere, so the contents of this block doesn't show up at the point of nesting (like footnotes) nor it is present in the wiki text (unlike the footnotes). Sorry for confusing description but the point is that plugins can do things to the instructions. And it is difficult to say in advance will those things get in the way if you try to recover the original wiki text from the instruction stream. So even if it sounds ugly, I still think that you could avoid some compatibility issues by taking offsets of links (or any other instructions of interest) from the instruction stream and then patching the text file at those offsets. Of course, it would be nice to verify that text at the offset matches the text from the instruction before you overwrite it with fixed link. And indeed regex's could be used for that ;) Mykola -- DokuWiki mailing list - more info at http://www.dokuwiki.org/mailinglist