Hacker News new | comments | show | ask | jobs | submit login
SQLite timeline items related to "json" (sqlite.org)
100 points by coleifer 763 days ago | hide | past | web | 22 comments | favorite



The JSON support in SQLite is in a very early stage of design and development still. There is a lot of cool stuff coming. But there is also still a lot of work to do.

Please check back in a few weeks.

If you would like to leave comments describing the kinds of problems you would like JSON support in SQLite to solve, I promise to read them all and give them careful consideration.


I don't care much for json data in tables, but I would really like to use aggregate functions returning json to create hierarchical results from queries with multiple levels of aggregation.


Id like to be able to compose subset JSON objects (give me only the following keys), be able to create indexes on values especially partial indexes where I can say something like x.likes > 5 or the like, some syntax for updating or reasoning about the object that is JSON-patch like.

I think in android this could be really important as a way to deflate objs into native types in c-land -- Java parsers on android are so slow that Facebook had to move to flat buffers to avoid them.


Just commenting to say thank you.


Hi, I spend at least $60 x 12 months /year on a NoSQL database that HN absolutely loves to hate. I don't mind the service, but my first love was SQLite (thanks to Android, and Python), and my second love was JSON (partly why I ended up choosing the database, and switched to the NoSQL database). Some of this is probably slightly off-topic, but since I love the product, I would very happily use SQLite once again if for:

1) JSON support, as is being offered as described in the post. I want to do the following in SQLite (look to MongoDB(!) documentation for syntax): a) Inserts. You know, regular inserts. b) Upserts. The 'hated' database (Mongo, obvs) has this easily-configurable upsert ['replacements'?], that's incredibly easy to use. c) Bulk upserts, inserts and deletes.

2) Cheap-ish online database that you can probably use Amazon S3 for, which I am going to make my institution (University/employer) pay for anyway. Things I care for are: a) you take care of all the deployment details and provide support when I need one and b) don't throw me off your service when I cross your limits because I've overused them because I'm a data-hoarder but I'll lose my life's worth of work if you throw me off your service.

I would also love it if your offered a choice of setting up couple of indexes very easily (using JSON, obvs), and offered a visual explorer of data in the database (relevant because a certain mongo service offers an awesome data explorer).

Please understand that a very large majority of your users is NOT doing complex joins or using other advances features of the database. As an engineer, I would love it if my users used the most advanced features and let me code the funnest, cutting-edge algorithms, but that's not what I want as a user. All I ask is simple inserts, deletes, upserts, an online service to handle them, and an ability to do this all using JSON. This is not going to make the most exciting days of your work-life, but I would be an extremely happy person if I were able to use these services.

Sidenote: Since I may never have chance to an SQLite person again, I'll bring this up even though this is likely WAY off-topic. I am a researcher who likes to HOARD data. I like storing data in my Uni's server, that has a limit on file size/total size. IF you created the ability to store the same database across several files, I would be most grateful to you.

PS: Love your documentation, guys. You really set the bar for the awesomest, most easily-understandable documentation of extremely technical and advanced subject matters in the field. All my attempts to make my code readable are attempts to merely emulate your awesomeness. : )


I doubt SQLite can ever be used on S3. There are two problems with that, first is that to preserve consistency SQLite requires working locking, which S3 by default can't provide (for sane reason using SQLite on NFS is not advised).

Second problem is that S3 does not allow to modify files, each time you make a change you need to reupload the whole thing.

I think you might have better luck trying DynamoDB.

Regarding the post, I don't know what to think. The SQLite was banned that way because it supposed to be light, I feel like adding things like JSON is going away from that principle.


"IF you created the ability to store the same database across several files, I would be most grateful to you."

http://www.sqlite.org/src/doc/trunk/src/test_multiplex.c


you cannot go wrong copying postgresql.


Oh hey, neat.

http://www.sqlite.org/src/artifact/950ed4e8deb8ad4c

So from looking at the test, you can get JSON out of select queries now, not a full blown JSON type like PostgreSQL.


The test is written in TCL, which has slightly non-obvious syntax. In essence it tests SQL functions json_array and json_object which seem to return JSON arrays and objects as string.


It's pretty straightforward to write a UDF that does exactly that actually, JSON type a-la Postgres would be amazing though.

https://gist.github.com/alixaxel/5562152#file-y-php-L212-L23...


If you are interested in mobile JSON databases, we've been working on our network-enabled native databases for iOS, Android, C# and PhoneGap, since 2010. The library size is less than 500kb and we are compatible with the Apache CouchDB sync protocol. We also have a server side Sync Gateway that has fine grained access control, webhooks, etc.

Did I mention everything is open source? There is a mini-hack here that will let you choose your platform and be up and running in ~20 minutes with what's basically the stem cell you're gonna be winning hackathons with now that you know.

https://github.com/couchbaselabs/mini-hacks

Ps we couldn't have done it without SQLite as we started out using that for our storage engine. Now we are moving to ForestDB because it's more like the data structures we present in our API.


Seems like a bad place to run a sales pitch.


It's a curious thing that Dr Richard Hipp co-created the "unql" specification then effectively did nothing more with it after that.

http://www.infoq.com/news/2011/08/UnQL


There's code, you can compile it and try it out. For fun I'm thinking to write a little python wrapper:

http://unql.sqlite.org/index.html/wiki?name=UnQL


Newbie here. Can somebody explain me what does the JSON SQL function extension? Any example?


At first glance it seems like some functions that like json_object which can return (query) results as a JSON array or json_insert to I presume puts JSON directly into SQL tables.

Nothing too spectacular and nothing you couldn't do with any decent library already, but some convenience.


Every big feature starts out small.


This is so exciting! To anyone who doubts the value of this effort - do you have any good option for an embedded, document-based DB system already? Thanks Dr Richard Hipp!


UnQLite. Not affiliated with SQLite, but its an embedded json document store. Supports a cool procedural query language.

http://charlesleifer.com/blog/introduction-to-the-fast-new-u...


I like it but I am just a bit skeptical about why there is an investment being done in building this.

Aren't they already too late to the party? I mean, there would have been tonnes of libraries which would do this given how popular SQLite is. Shouldn't they have left it to those libraries and stuck to investing on the core database thingy ( too vague but I hope you get what I mean ).

:)


I'd rather have this built in SQLite, and with its standard code quality and extensive testing, than have to use some additional library.

And what does "too late to the party" even mean? Who said the party ended or ever ends?




Applications are open for YC Winter 2018

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact

Search: