Jan Hidders wrote: I have indeed been told by a friend that it would be a better idea to write down the BNF for the grammar before starting to write the yacc file. So I tried. But then, when I finally did the yacc file, I realised that it was not such a good idea after all. In fact, it was a waste of time. You need to be able to test the grammar while you are working on it, and you want to be able to tweak certain parts before writing certain other parts that depend on them. This is quite different from planning a software project, where you sketch your system and its modules in a very rough outline first. Rather, when writing the yacc file, you are actually writing the grammar. With BNF, you would also be writing the grammar, but without the benefit of being able to test it. In particular, in BNF you wouldn't be able to described transformations like the one I needed to do for lists. You would instead rack your brains over how to parse the lists directly, never coming to a conclusion (because it is probably impossible). If you are still interested, please find below the BNF I originally wrote. However, the current yacc file I use is quite different from it. I haven't even done links yet, and I already know that the below grammar will not work for links, because it will treat the pipe character as text inside the double square brackets. Timwi article ::= e | block | article NEWLINE+ block block ::= paragraph | heading | bulletedlist | numberedlist | preblock | comment paragraph ::= nonemptytext | paragraph NEWLINE nonemptytext comment ::= EMPTYCOMMENT | BEGINCOMMENT TEXT ENDCOMMENT | BEGINCOMMENT TEXT | BEGINCOMMENT heading ::= headinglevel1 | headinglevel2 | headinglevel3 | headinglevel4 | headinglevel5 | headinglevel6 headinglevel1 ::= HEADING1 nonemptytext ENDHEADING? headinglevel2 ::= HEADING2 nonemptytext ENDHEADING? headinglevel3 ::= HEADING3 nonemptytext ENDHEADING? headinglevel4 ::= HEADING4 nonemptytext ENDHEADING? headinglevel5 ::= HEADING5 nonemptytext ENDHEADING? headinglevel6 ::= HEADING6 nonemptytext ENDHEADING? bulletedlist ::= LISTBULLET text | bulletedlist NEWLINE LISTBULLET text | bulletedlist NEWLINE NEWLINE LISTBULLET text numberedlist ::= LISTNUMBERED text | numberedlist NEWLINE LISTNUMBERED text | numberedlist NEWLINE NEWLINE LISTNUMBERED text preblock ::= PRELINE text | preblock NEWLINE* PRELINE text nonemptytext ::= intext | intext nonemptytext text ::= e | nonemptytext intext ::= TEXT | EXTENSION | comment | dblsqbr | PIPE | templateinclusion | templatevariable | bold | italics | APOSTROPHE | DOUBLEQUOTE # Things that are invalid syntax but we need to allow: | ENDCOMMENT | OPENDBLSQBR CLOSEDBLSQBR | OPENLINK CLOSEDBLSQBR dblsqbr ::= linkorimageorcategory | forcedlink linkorimageorcategory ::= OPENDBLSQBR linkinside CLOSEDBLSQBR forcedlink ::= OPENLINK linkinside CLOSEDBLSQBR linkinside ::= nonemptytext | linkinside PIPE nonemptytext

Jan Hidders wrote: Yes, I know that this is what you meant. But personally I think it is easier to write the parser first, and then use the yacc source file as the grammar description. Of course, one obvious caveat of this is that the yacc file will also contain grammar rules whose purpose is to catch invalid markup. This can be seen as a good thing, though: it means the grammar describes even what happens exactly when invalid markup is entered. Timwi