GNOME Bugzilla – Bug 688932
GJSON: New module to read/write JSON from G{Input,Output}Stream
Last modified: 2016-12-06 18:56:21 UTC
While porting gnome-ostree from python to gjs, I have JSON stored in files, but no way to read or write them. This module acts as a bridge between GIO and the native SpiderMonkey JSON. It's not very efficient since there's a lot of bouncing around of UTF-8 -> UTF-16 -> UTF-8, but, such is life with gjs.
Created attachment 229724 [details] [review] GJSON: New module to read/write JSON from G{Input,Output}Stream
(In reply to comment #0) > While porting gnome-ostree from python to gjs, I have JSON stored in > files, but no way to read or write them. var data = GLib.file_load_contents(filename); var obj = JSON.parse(data);
(In reply to comment #2) > (In reply to comment #0) > > While porting gnome-ostree from python to gjs, I have JSON stored in > > files, but no way to read or write them. > > var data = GLib.file_load_contents(filename); > var obj = JSON.parse(data); I second Jasper here. If data is not usable as a string (assuming it is good utf8), that's a bug in our byte arrays, not something to work around with a native module.
(In reply to comment #2) > (In reply to comment #0) > > While porting gnome-ostree from python to gjs, I have JSON stored in > > files, but no way to read or write them. > > var data = GLib.file_load_contents(filename); > var obj = JSON.parse(data); But that's really inefficient; any time you find yourself copying the entire contents of a file into memory, you're doing it wrong.
(In reply to comment #4) > (In reply to comment #2) > > (In reply to comment #0) > > > While porting gnome-ostree from python to gjs, I have JSON stored in > > > files, but no way to read or write them. > > > > var data = GLib.file_load_contents(filename); > > var obj = JSON.parse(data); > > But that's really inefficient; any time you find yourself copying the entire > contents of a file into memory, you're doing it wrong. Also, it doesn't give any story for how you'd parse JSON from other sources, like a socket or pipe.
What is Mozilla doing for their JSON story? node.js? Is there any prior art we can stare at? Streaming data like this is going to be more prevalent in the future with JavaScript, so I hope future editions of ECMAScript will tackle it.
(In reply to comment #6) > What is Mozilla doing for their JSON story? node.js? Well...a JS-native JSON API (named JSON) was added in ES5, relatively recently in 2009; it just parses strings. Internally, spidermonkey does have an incremental parsing API. > Is there any prior art we > can stare at? The aforementioned js185 API. It'd certainly be possible to expose it to JS..would gain in efficiency in some ways, lose in others. > Streaming data like this is going to be more prevalent in the future with > JavaScript, so I hope future editions of ECMAScript will tackle it. Well, pure JS has no streams that I know of. But yes. So, options for this patch: 0) Apply 1) Skip it, people parsing JSON with gjs can just load entire files into memory for now. I'm not opposed to that. 2) Write a new module which exposes the already extant incremental parser to JS, allowing people to feed data from whatever they like (including GInputStream)
For a streaming parser, we have imports.gi.Json.JsonParser nowadays.