Memoize lookup in parse is producing false positives

[henrinormak]

This issue relates to the use of memoization in this code: https://github.com/i18next/i18next-icu/blob/master/src/index.js#L47-L59

Namely, given two strings with the following keys, foo.bar.baz and foo.bar, if parse happens to run first for foo.bar.baz, the later lookup for foo.bar will produce a false positive (the check on line 54 is truthy, eventhough this particular key has not yet been memoized), which leads to an error at line 58 (fc.format is undefined and not a function).

Likely it would be enough to swap the simple truthy check with something a bit more accurate (such as !fc && typeof fc.format === 'fuction').

Until then, the only real workaround (without changing keys) is to disable memoization (which is an overhead). Please let me know if you need further details, I can file a PR if needed.

[henrinormak]

I suppose the memoization could perhaps work differently as well, avoiding nesting and instead using the key as a concrete property (i.e changing the way the utils work), but as I lack the context I might be missing something on why the memoize memory needs to be a deep object. Performance-wise it might actually be better to have a flat object.

It also raises a question about keys that contain .format. in them, as those would potentially have a conflict with the memoized IntlMessageFormat object.

[jamuhl]

how can you have 2 keys foo.bar and foo.bar.baz existing side by side? Did you set keySeparator to false on init? If so yes...we need to fix this by changing that to replace the . in keys (as that is the fallback on that set/get) https://github.com/i18next/i18next-icu/blob/master/src/index.js#L52

fc = utils.getPath(this.mem, ${lng}.${ns}.${key.replace('.', '###')});

would you like to provide a PR?

[henrinormak]

Yes, I have configured keySeparator: false in i18next.init, I'll open a PR to do the replace.