Connector for browser indexedDB

[bajtos]

At the moment, loopback's memory connector can be configured to persist data in browser's local storage. This has a significant problem as the local storage is size limited to about 5MB.

We should support storing the data in indexdb too, possibly via https://github.com/google/lovefield

A community-run implementation: https://github.com/r3dm/loopback-connector-indexedDB

[vigyanhoon]

5MB is practically too less for a major application housing all data offline.

Please add IndexedDB support. We can't think of an application to survive only on 5 MB where most part of the app is going to be offline.

[ffflabs]

How is localstorage support provided? I thought memory connector was persisted server side.

[bajtos]

See Memory.prototype.loadFromFile and Memory.prototype.saveToFile.

We had extended memory connector to support both fs files and browser's localStorage.

It is possible to extend the memory connector to load/save data in indexeddb, but such solution would be be very memory hungry, as the page will load whole indexeddb into javavascript memory (heap). A better solution is to implement a new connector that will fetch the data from the indexeddb on per-record basis as needed.

[jrschumacher]

What about using Mozilla's localForge solution?

[bajtos]

A powerful solution would be to implement a levelup connector, which could be then configured to use either LevelDB on the server or IndexedDB in the browser (via level.js).

However, both modules are quite large and thus the final client bundle would be even bigger than it is now.

In the light of that, I am inclining towards writing a connector directly on top of the IndexedDB API, so that the module size is as small as reasonably possible.

[jrschumacher]

Ah, that makes sense. Maybe it could be a wrapper which try's to detect leveldb or local forge and if found then resorts to the included library?

[bajtos]

Maybe it could be a wrapper which try's to detect leveldb or local forge and if found then resorts to the included library?

API wise, we could use localForge instead of IndexedDB. However, [https://mozilla.github.io/localForage/localforage.min.js] is almost 24kb, which is a problem.

Why are you pushing for localForage, what are the advantages compared to pure IndexedDB?

[jrschumacher]

I'm not. Just what I know.

[bajtos]

I'm not. Just what I know.

Fair enough. I was wondering if there is something important that I am missing.