Re: extension to string.gmatch (structural regular expressions)

["Duncan Cross" <duncan.cross@...>]

On Jan 3, 2008 11:50 AM, Shmuel Zeigerman <shmuz@actcom.co.il> wrote:
> Miles Bader wrote:
> >
> > Though in many cases, you don't really need split, e.g. something like
> > the following is actually cleaner than doing a split and iterating over
> > the resulting table:
> >
> >    for field in string.gmatch (STRING, "[^ \t\n]+") do
> >      .... use field ...
> >    end
>
> And what if the fields are delimited by, say, "<delim>" ?
>
> The rule of thumb is to use gmatch when fields are easier to express
> with regexp than delimiters, otherwise use split (provided it's
> available :) ).
>
> --
> Shmuel
>

Here's a first attempt at a splitting gmatch:

function string.gmatchsplit(text, pattern)
    local gm = text:gmatch("()" .. pattern .. "()")
    local start = 1
    local function aux(...)
        if start == nil then
            return nil
        end
        local caps = select('#', ...)
        if caps == 0 then
            local finalpart = text:sub(start)
            start = nil
            return finalpart
        end
        local stop = (...) - 1
        local before = text:sub(start,stop)
        start = select(caps, ...)
        return before, select(2, ...)
    end
    return coroutine.wrap(
        function()
            while true do
                coroutine.yield( aux(gm()) )
            end
        end
    )
end

for part in string.gmatchsplit("one<delim>two<delim>three", "<delim>") do
   print(part)
end