The logic

if (Object.prototype.hasOwnProperty.call(parent, peer) &&
            parent[peer] !== undefined) {

is repeated 4 times in lib/types/object/index.js as well as

if (!Object.prototype.hasOwnProperty.call(parent, peer) ||
            parent[peer] === undefined) {

being repeated twice.

Does it really only make sense to support the nested syntax only for .or() given so many other methods share (and could benefit from) the same syntax? #1553 was another request for nested paths in additional methods.

I am eagerly awaiting this feature for .xor(), so I was a little disappointed to see that this PR only resolved this problem for .or().

Don't worry I won't merge it until all are supported, baby steps… 🙂

I'm working on it. I'll implement the change for other methods.

All done! @Marsup @dhm116

Kindly review!

This looks great - exactly what I was hoping for!

The only concern I would have, which comes from my own fiddling, would be the incompatibilities that arise around scenarios regarding the usage of labels, like in the it('should apply labels'... tests from the with(), without(), or(), and() and nand() scenarios.

I believe with the current implementation, if you duplicate the should apply labels tests but use the nested peer syntax, the test would fail - I could be wrong, it was pretty late at night!

True @dhm116 ... Just tried it out now and the tests failed.

I haven't used labels before but I'll look into its usage and see how to fix this.

It's tricky with labels for nested objects. I'd appreciate some help here.

This probably needs to be cleaned up a bit, but this seems to pass the tests now - BolajiOlajide/joi@ft-add-support-for-nested-objects...dhm116:ft-add-support-for-nested-objects

internals.keysToLabels = function (schema, keys) {

    const children = schema._inner.children;

    if (!children) {
        return keys;
    }

    const findNestedLabel = function (key) {

        const [parentKey, ...parts] = key.split('.');
        const childKeys = parts.join('.');
        const matchingChild = children.find((child) => child.key === parentKey);
        const label = matchingChild ? internals.keysToLabels(matchingChild.schema, childKeys) : key;

        return !key.endsWith(label) ? label : key;
    };

    const findLabel = function (key) {

        if (key.includes('.')) {
            return findNestedLabel(key);
        }
        const matchingChild = children.find((child) => child.key === key);
        return matchingChild ? matchingChild.schema._getLabel(key) : key;
    };

    if (Array.isArray(keys)) {
        return [].concat(keys.map(findLabel));
    }

    return findLabel(keys);
};

You could either clean this up (after @Marsup weighs in) and update your PR, or I can submit a PR to your branch.

Let's wait for @Marsup to weigh in. Once he's approved, A PR submitted to my branch would be great.

Try that see if it works:

--- i/lib/types/object/index.js
+++ w/lib/types/object/index.js
@@ -767,8 +767,8 @@ internals.keysToLabels = function (schema, keys) {
 
     const findLabel = function (key) {
 
-        const matchingChild = children.find((child) => child.key === key);
-        return matchingChild ? matchingChild.schema._getLabel(key) : key;
+        const matchingChild = schema._currentJoi.reach(schema, key);
+        return matchingChild ? matchingChild._getLabel(key) : key;
     };
 
     if (Array.isArray(keys)) {

Thanks @Marsup . That works.

I'll refactor and add more tests and push in a few minutes.

I've submitted https://github.com/BolajiOlajide/joi/pull/1 based upon @Marsup's beautiful simplification, which produces the following test results:

1123 tests complete
Test duration: 3341 ms
Assertions count: 21153 (verbosity: 18.84)
No global variable leaks detected
Coverage: 100.00%
Linting results: No issues

Hey @Marsup ,

I've merged @dhm116's PR, I think this PR is good now.

It looks alright, just missing the docs. I changed the base branch as it's going to be a breaking change.

What do I need to do @Marsup ?

Actually let me think about that one, it's looking more and more like we could refs instead of plain strings.

Hey @Marsup ,
What's the status of this PR?

Thanks a lot!

This thread has been automatically locked due to inactivity. Please open a new issue for related bugs or questions following the new issue template instructions.