News for hackers. Stuff that matters.
Re:What is the most intrusive wart on your favourite program... by bayareaguy on Monday December 08, 2008 @02:40AM (#389626)
I use Lua a lot these days and coming from Python, it seems silly that I have to provide my own empty? function since standard Lua doesn't offer any way to get the number of elements in an arbitrary table other than to iterate over it:

    function empty_p(t)
        for k,v in pairs(t) do
            return true
        end
        return false
    end


    -- later on
    if not empty_p(t) then
        do_processing(t)
    end
This is because although Lua offers a #t operator, it's only guaranteed to be correct if you're careful about how the table is constructed. That and the weird ipairs() behavior for non-trivial tables drives me crazy sometimes. Here's an example:

    Lua 5.1.2  Copyright (C) 1994-2007 Lua.org, PUC-Rio
    > t = {1,2,3}
    > print(#t)
    3
    > for i,v in ipairs(t) do print(i,v) end
    1       1
    2       2
    3       3
 
Ok so all is good. But now what happens when the table contains a nil?

    > t = {1,nil,3}
    > print(#t)
    3
    > for i,v in ipairs(t) do print(i,v) end
    1       1
ipairs() stops at the first nil value even though Lua knows there is data past it. Things get more interesting for tables with non-integer keys:

    > t = {}
    > t['a'] = 1
    > t['b'] = 2
    > print(#t)
    0
    > t[1] = 1
    > print(#t)
    1
    > for i,v in ipairs(t) do print(i,v) end
    1       1
What's going on here is each Lua table has an array for integer valued keys and a separate hash table for non-integer keys and # and ipairs() only operate on the first array. To process all the keys you use pairs():

    > for k,v in pairs(t) do print(k,v) end
    1       1
    a       1
    b       2
While at first this doesn't seem to be such a big issue, it's easy to forget when it's safe to use the #t construct. I've made this error so often that now whenever my Lua code doesn't do what I expect the first thing I look for are places where I have something like:

    if #t > 0 then
        do_processing(t)
    end
or

    for i,v in ipairs(t)
        process(v)
    end
because chances are the real problem will turn out to be the particular way I initialized the table.
Re:What is the most intrusive wart on your favourite program... by silentbicycle on Monday December 08, 2008 @11:31AM (#389980)
I suspect this is because a table may be used for an object (or some other abstract data type) which has a different data type. Adding another metatable hook for __empty might be another solution, though. (I agree it's a bit annoying.)

> Ok so all is good. But now what happens when the table contains a nil?

One of the invariants for the way Lua tables are implemented is that they won't contain nil as a value, IIRC. Using a multifaceted table type for both arrays and dicts is a mixed blessing. It is really convenient much of the time, but it does lead to some quirks.

NB: The i in "ipairs" means "integer (keyed) pairs".