Accepts empty string as valid semver (i.e. treats major component as optional)

[JC3]

I don't think it's generally a problem to be loose with semver parsing, but I wonder if semver_parse("", ...) should fail instead of the current behavior of yielding 0.0.0?

A related side effect of this is "-1.2.3" is parsed as 0.0.0 with prerelease "1.2.3", which probably wasn't the intent of that string. In general, anything that forms a valid prerelease and/or metadata suffix is accepted as a valid semver in itself.

And other strangeness:

void testparse (const char *str) {

    semver_t v = {};

    printf("%-10s: ", str);
    if (semver_parse(str, &v))
        printf("error");
    else
        printf("major=%d  minor=%d  patch=%d  pre=%-8s  meta=%s", 
               v.major, v.minor, v.patch, v.prerelease, v.metadata);
    printf("\n");

    semver_free(&v);

}

int main () {

    testparse("");
    testparse("-1.2.3");

    testparse("-");
    testparse("+");
    testparse("-+");
    testparse("-3.14159");
    testparse("+10");

    testparse("fail");

}

Outputs:

          : major=0  minor=0  patch=0  pre=(null)    meta=(null)
-1.2.3    : major=0  minor=0  patch=0  pre=1.2.3     meta=(null)
-         : major=0  minor=0  patch=0  pre=          meta=(null)
+         : major=0  minor=0  patch=0  pre=(null)    meta=
-+        : major=0  minor=0  patch=0  pre=          meta=
-3.14159  : major=0  minor=0  patch=0  pre=3.14159   meta=(null)
+10       : major=0  minor=0  patch=0  pre=(null)    meta=10
fail      : error

All of those seem like they should, at least philosophically, result in parse failures. 🤷‍♂️

Having the parser verify that a major component is present would cover all those cases as well as the empty string case.

I mean, there's loose, and then there's loose... 😂