Question

I'm currently getting familiar with the LPeg parser module. For this I want to match a version string (e.g. 11.4) against a list.

Such a list is a string with a tight syntax that can also contain ranges. Here is an EBNF-like, but in any case quite simple grammar (I write it down because LPeg code below can be a bit difficult to read):

S = R, { ',', R }
R = N, [ '-', N ]
N = digit+, [ '.', digit+ ]

An example string would be 1-9,10.1-11,12. Here is my enormous code:

local L = require "lpeg"
local LV, LP, LC, LR, floor = L.V, L.P, L.C, L.R, math.floor
local version = "7.25"

local function check(a, op, b)
    if op and a+0 <= version and version <= b+0 then
        return a..op..b -- range
    elseif not op and floor(version) == floor(a+0) then
        return a        -- single item
    end
end
local grammar = LP({ "S",
    S = LV"R" * (LP"," * LV"R")^0,
    R = LV"V" * (LC(LP"-") * LV"V")^-1 / check,
    V = LC(LV"D" * (LP"." * LV"D")^-1),
    D = (LR("09")^1),
})
function checkversion(str)
    return grammar:match(str)
end

So you would call it like checkversion("1-7,8.1,8.3,9") and if the current version is not matched by the list you should get nil.

Now, the trouble is, if all calls to check return nothing (meaning, if the versions do not match), grammar:match(...) will actually have no captures and so return the current position of the string. But this is exactly what I do not want, I want checkversion to return nil or false if there is no match and something that evaluates to true otherwise, actually just like string:match would do.

If I on the other hand return false or nil from check in case of a non-match, I end up with return values from match like nil, "1", nil, nil which is basically impossible to handle.

Any ideas?

Was it helpful?

Solution 2

This is the pattern I eventually used:

nil_capturing_pattern * lpeg.Cc(nil)

I incorporated it into the grammar in the S rule (Note that this also includes changed grammar to "correctly" determine version order, since in version numbering "4.7" < "4.11" is true, but not in calculus)

local Minor_mag = log10(Minor);
local function check(a, am, op, b, bm)
    if op then
        local mag = floor(max(log10(am), log10(bm), Minor_mag, 1))+1;
        local a, b, v = a*10^mag+am, b*10^mag+bm, Major*10^mag+Minor;

        if a <= v and v <= b then
            return a..op..b;
        end
    elseif a == Major and (am == "0" or am == Minor) then
        return a.."."..am;
    end
end

local R, V, C, Cc = lpeg.R, lpeg.V, lpeg.C, lpeg.Cc
local g = lpeg.P({ "S",
    S = V("R") * ("," * V("R"))^0 * Cc(nil), 
    R = (V("Vm") + V("VM")) * (C("-") * (V("Vm") + V("VM")))^-1 / check,
    VM = V("D") * Cc("0"),
    Vm = V("D") * "." * V("D"),
    D = C(R("09")^1),
});

OTHER TIPS

I think you can or + it with a constant capture of nil:

grammar = grammar + lpeg.Cc(nil)

Multiple returns from match are not impossible to handle, if you catch them in a way that makes handling them easier. I added a function matched that does that, and added the fallback return of false to your check.

do
    local L = require "lpeg"
    local LV, LP, LC, LR, floor = L.V, L.P, L.C, L.R, math.floor
    local version = 6.25

    local function check(a, op, b)
        if op and a+0 <= version and version <= b+0 then
            return a..op..b -- range
        elseif not op and floor(version) == floor(a+0) then
            return a        -- single item
        end
        return false
    end
    local grammar = LP({ "S",
        S = LV"R" * (LP"," * LV"R")^0,
        R = LV"V" * (LC(LP"-") * LV"V")^-1 / check,
        V = LC(LV"D" * (LP"." * LV"D")^-1),
        D = (LR("09")^1),
    })

    local function matched(...)
        local n = select('#',...)
        if n == 0 then return false end
        for i=1,n do
            if select(i,...) then return true end
        end
        return false
    end

    function checkversion(ver,str)
        version = ver
        return matched(grammar:match(str))
    end
end

I enclosed the whole thing in do ... end so that the local version which is used here as an upvalue to check would have constrained scope, and added a parameter to checversion() to make it clearer to run through few test cases. For example:

cases = { 1, 6.25, 7.25, 8, 8.5, 10 }
for _,v in ipairs(cases) do
    print(v, checkversion(v, "1-7,8.1,8.3,9"))
end

When run, I get:

C:\Users\Ross\Documents\tmp\SOQuestions>q18793493.lua
1       true
6.25    true
7.25    false
8       true
8.5     true
10      false

C:\Users\Ross\Documents\tmp\SOQuestions>

Note that either nil or false would work equally well in this case. It just feels saner to have collected a list that can be handled as a normal Lua array-like table without concern for the holes.

Licensed under: CC-BY-SA with attribution
Not affiliated with StackOverflow
scroll top