The other day I was converting the output of a Git::Raw::Commit into JSON using my module JSON::Create, when I noticed an oddity:

{
"commits":[
    {
        "body":null,
        "id":"27ed4669e32ce2d14831c719dfd5b341a659788e",
        "message":"Remove a stray html ending tag\n",
        "time":"1609997818"
    },

The "time" part always comes out as a string, even though it's clearly a number. Was this a bug in my module, some kind of dual-string-and-number wannabee variable which JSON::Create falsely turned into a string?

As it happens, no. Git::Raw actually puts the number into a string. (The newSVpv there makes a new Perl string, and the sprintf above that does exactly the same job as Perl's sprintf.)

So Git::Raw turns the original C variable of the form git_time_t, a 64-bit integer type representing the number of seconds since the "epoch" (1970), into a string, perhaps to avoid the "year two million" problem or whatever, because Perl can hold up to 52 or 53 bit integers.

Anyway Perl's monkey business with numbers and strings, and the lack of booleans, makes creating JSON quite complicated, although not as complicated as identifying cats in photographs and youtube videos.

One thing which I like about Mojolicious is that they put all the functions and methods in alphabetical order. When you get used to that then go back to a module like Git::Raw::Repository with a large number of functions in apparently random order, it does seem like quite a smart move for the reader to use the alphabetical ordering.

Anyway I thought so, so I've been putting all the functions, methods, and other things in my modules into alphabetical order. I even started to write tests that they are all alphabetical, since I usually manage to slip up on these things.