-
Notifications
You must be signed in to change notification settings - Fork 9
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
"Strict mode" to disallow additional fields? #15
Comments
It would indeed be possible. I'm not 100% sure you'd want strict parsing for the whole document though, but rather be more granular about it: typedef Strict<T> = T;
typedef FlatStrict<T> = T; //strict about T, but not recursively
typedef Lenient<T> = T; Or maybe there are better ways, I don't really know. My main motivation for tink_json was to consume data from external sources, so I did want proper errors if some fields were changed / removed, but skip any additions - or data that doesn't interest me to start with. So I don't really have any opinion on this. What do you think about more granularity? |
More granularity sounds good in principle, and that typedef-approach seems flexible enough. A more general note: I was looking into tink_json for parsing haxe-formatter config files (very much WIP btw, not really usable yet :D). However, it seems to be more performance-oriented, and I don't care about performance so much for parsing a single json file, just about having error messages that are as nice as possible. I think json2object will be a better fit for that (once it supports abstracts). :) I like how json2object handles additional fields btw: it puts a warning into |
Well, concerning your use case, I would caution against warning user about superfluous fields in tooling configs. It is very common practice for tools that are build on top to add fields into the basic config (just consider what some tools do to npm's As for the rest, tink_json happens to be fast (in fact according to @nadako on Haxe 3.4 it's faster on JS than native JSON parsing), but performance is merely a secondary goal. If you look closer, you will discover that this library does far more than json2object. But I really don't want to waste your time with weird attempts to sell it to you. If there's a lib that tackles the subset you care about in a manner that suits you more, then that's awesome. I just wish they hadn't built it entirely from scratch ... but where would that leave all the fun, right? ^^ |
Hm... I think it's more important to catch typos and the like, improving the user experience (which affects every user), than to support tools built on-top (which only affects some, and only in case such tools are made). Those tools should be able to manage by only passing what's needed to the formatter, or if needed, there could be a Anyway, this is getting a bit off-topic. Thanks for the advice. :) |
Would it be possible to have a "strict mode" that errors when the parser encounters fields that are not part of the specified type? The "Reading" section in the Readme states:
So it seems additional fields are just ignored right now.
In a "strict mode", the parser should error with
Unexpected field 'f'
in this example:The text was updated successfully, but these errors were encountered: