WASM: An option to disable "Installing sqlite3 bits"
(1) By Daniel Steigerwald (steida) on 2023-01-31 19:38:56 [link] [source]
Is this the right place to request features? Anyway, while S is probably useful, it would be great if it could be disabled. Thank you
(2) By Stephan Beal (stephan) on 2023-02-01 01:49:53 in reply to 1 [link] [source]
Is this the right place to request features?
It is.
Anyway, while S is probably useful, it would be great if it could be disabled
The global S should only be enabled in localhost mode. Are you seeing it elsewhere? There's no really strong argument against removing it from the core, instead requiring each app to export it if they want.
(3) By Daniel Steigerwald (steida) on 2023-02-01 22:43:57 in reply to 2 [link] [source]
I am asking for this feature because my library embeds Sqlite WASM, and I don't want to confuse users with this console.log.
(4) By Stephan Beal (stephan) on 2023-02-02 03:34:00 in reply to 3 [link] [source]
I am asking for this feature because my library embeds Sqlite WASM,
Fair enough. i'll get it removed in the next day or two.
(5) By Daniel Steigerwald (steida) on 2023-02-02 15:09:58 in reply to 4 [link] [source]
Thank you :-)
(6.1) By Daniel Steigerwald (steida) on 2023-02-06 13:26:29 edited from 6.0 in reply to 4 [link] [source]
"Ignoring inability to install OPFS sqlite3_vfs:"
Another console.warn I would rather not show it to users of my library. I suppose only real, unavoidable errors should be shown. For debugging, I suppose there can be some location query parameter like "?enable_sqlite_warnings" or explicit dev build. What do you think?
(7) By Stephan Beal (stephan) on 2023-02-06 13:39:11 in reply to 6.1 [link] [source]
"Ignoring inability to install OPFS sqlite3_vfs:" ... What do you think?
i'm torn on that one primarily because of the support overhead i will have for not having that output to refer to when people post here saying "my OPFS doesn't work" (that has been the single most-posted query about js/wasm to date).
i agree wholeheartedly with you that the library should be silent when being embedded within another layer of library but we also need a way for folks to easily tell why OPFS is not loading for them. Providing a URL argument to enable that still requires a round-trip to the forum where someone will post "it doesn't work" and i'll have to answer "try again with param so-and-so, added here, here, or here, depending on how you're loading it..." (Let's face it - most people don't read the docs!)
There is a "back door" in place for feeding in configuration options, which we "could" use for this. Would the following be too onerous for your particular case?
self.sqlite3ApiConfig = {silent: true};
import sqlite3 ....;
const sqlite3 = await sqlite3ApiInit();
That's hypothetical code - the "silent" flag is not implemented but the bootstrap function does look for a global named sqlite3ApiConfig as a way to slip it config options during the loading/bootstrapping phase. i.e. this would be a simple way to implement this, in particular if the above is hidden from your clients.
:-?
(8) By Daniel Steigerwald (steida) on 2023-02-06 13:46:20 in reply to 7 [link] [source]
The approach I would use is to make API explicit. So instead of warnings, I would explicitly throw an error when a developer tries to make an impossible instance - new sqlite3.oo1.OpfsDb("/evolu/evolu1.db", "c")
Then throwing an error with an explanation makes sense to me.
My library chooses OPFS/localStorage explicitly by navigator.userAgentData
so I can control when I enable the OPFS version.
(9) By Stephan Beal (stephan) on 2023-02-06 14:11:39 in reply to 8 [source]
would explicitly throw an error when a developer tries to make an impossible instance - new sqlite3.oo1.OpfsDb("/evolu/evolu1.db", "c")
It is made impossible in the sense that that class is not installed unless OPFS is available. i'd much rather than the class simply not be there if OPFS is not available because it gives clients an easy way to query whether OPFS is an option or not before attempting to use it. e.g. via something like:
const MyDb = sqlite3.oo1.OpfsDb || sqlite3.oo1.DB;
const db = new MyDb(...);
i've just checked in a change in which the library now uses configurable console.log/debug/warn/error() workalikes which can be swapped out by clients. To extend the earlier example:
self.sqlite3ApiConfig = {warn: ()=>{}}; // squelch all warnings
import ...;
delete self.sqlite3ApiConfig;
Whether that's sufficient for your use case, i'm not sure, but it has other uses so it was a reasonable change to make anyway.
i'd currently really rather not disable that warning globally, for all clients, because of the inherent forum-side support burden doing so will impose on me personally. That's not to say that my mind won't change/can't be changed, though.
i'm heading out for the rest of the day so won't be responsive but am interested in finding a solution to this.
(10) By Daniel Steigerwald (steida) on 2023-02-06 14:36:32 in reply to 9 [link] [source]
Thank you. Could you please make a prerelease so that I can test the changes?
(11) By Stephan Beal (stephan) on 2023-02-06 14:39:49 in reply to 10 [link] [source]
Could you please make a prerelease so that I can test the changes?
A snapshot was just uploaded: wasm:/uv/snapshot.html
(12) By Daniel Steigerwald (steida) on 2023-02-06 20:18:10 in reply to 11 [link] [source]
Uncaught ReferenceError: sqlite3 is not defined at installAsyncProxy (sqlite3-opfs-async-proxy.js?c676:74:12) at eval (sqlite3-opfs-async-proxy.js?c676:878:5) at ../../packages/evolu/dist/sqlite/sqlite3-opfs-async-proxy.js (packages_evolu_dist_sqlite_sqlite3-opfs-async-proxy_js.js:20:1) at options.factory (packages_evolu_dist_sqlite_sqlite3-opfs-async-proxy_js.js:541:31) at webpack_require (packages_evolu_dist_sqlite_sqlite3-opfs-async-proxy_js.js:50:33)
I recommend Next.js for testing https://nextjs.org/docs/getting-started
(13) By Stephan Beal (stephan) on 2023-02-06 20:23:21 in reply to 12 [link] [source]
(sqlite3-opfs-async-proxy.js?c676:74:12) at eval (sqlite3-opfs-async-proxy.js?c676:878:5) at
Stupid user error on my part - will be resolved when i'm back on a computer. That file runs in a dedicated worker and doesn't have access to the sqlite3 object.
(14) By Stephan Beal (stephan) on 2023-02-06 21:25:33 in reply to 13 [link] [source]
That file runs in a dedicated worker and doesn't have access to the sqlite3 object.
That's been corrected and a new snapshot uploaded. That particular file still can use console.debug/warn/error but "shouldn't" ever do so unless its verbosity level is increased via an explicit URL argument.
(15) By Daniel Steigerwald (steida) on 2023-02-10 22:01:05 in reply to 14 [link] [source]
For some reason, I still see those warnings when I'm using LocalStorage.
(16) By Stephan Beal (stephan) on 2023-02-10 22:04:42 in reply to 15 [link] [source]
For some reason, I still see those warnings when I'm using LocalStorage.
Can you post the exact message? The one about installing the global "S" was moved from the library into the main test app, so it should only be showing up there.
PS: thanks for the FF/Safari updates.
(17) By Daniel Steigerwald (steida) on 2023-02-11 20:28:35 in reply to 16 [link] [source]
Run https://www.evolu.dev/ in Firefox or Safari.
It shows: Ignoring inability to install OPFS sqlite3_vfs: The OPFS sqlite3_vfs cannot run in the main thread because it requires Atomics.wait().
(18) By Stephan Beal (stephan) on 2023-02-12 10:42:43 in reply to 17 [link] [source]
It shows: Ignoring inability to install OPFS sqlite3_vfs: The OPFS sqlite3_vfs cannot run in the main thread because it requires Atomics.wait().
Good morning,
That one is covered by the new ability to override the warn/debug output routines before the library is ititialized. If you'll do this before your "import", all such messages will go away:
self.sqlite3ApiConfig = { debug: ()=>{}, log: ()=>{}, warn: ()=>{}, error: ()=>{} }
This was just tested on Firefox, and the above addition silences any and all debug/warning/error output emitted by the library-level code (as opposed to test app code).
After the import you can delete self.sqlite3ApiConfig
- it's only evaluated (and any parts which are needed are copied) during the initial code import and bootstrapping process.
Though i'm 100% willing to remove almost all forms of warning output from the lib by default, the various "cannot load the OPFS VFS for reason so-and-so" warnings exist to reduce the support burden and back-and-forth here in the forum, as "why doesn't my OPFS work?" has been the single most-asked question so far and seems likely to remain so. The output overrides demonstrated above are the current compromise for that.
(19) By Daniel Steigerwald (steida) on 2023-02-12 20:12:36 in reply to 18 [link] [source]
Thank you, that works.