sql.js
sql.js copied to clipboard
sql-js
Multiple connection support?
Is there any way to model multiple connections to the same DB in sql.js (with whatever transaction isolation between them), or any way to fake the same effect?
Nice idea. I would suggest asking on the SQLite mailing list if there could be any way to support multiple connections to a memory database (it would actually be nice for me as well).
Alternative could be some kind of a virtual shared memory database or "ramdisk". I found some interesting here: https://nedbatchelder.com/blog/201105/running_a_database_from_a_ramdisk.html
mm, thanks for the hint. Looks like https://www.sqlite.org/inmemorydb.html can indeed support multiple connections, either with or without shared cache. However, I'm not sure that sql.js is using in-memory? From looking at api.coffee, it seems to create a fake dbfile of some kind:
class Database
# Open a new database either by creating a new one or opening an existing one,
# stored in the byte array passed in first argument
# @param data [Array<Integer>] An array of bytes representing an SQLite database file
constructor: (data) ->
@filename = 'dbfile_' + (0xffffffff*Math.random()>>>0)
if data? then FS.createDataFile '/', @filename, data, true, true
@handleError sqlite3_open @filename, apiTemp
...which doesn't look like it exposes the ability to open the file multiple times. Conversely, it looks like it might be fairly easy to add...
...completely untested, i think something like this could work, by expanding Database() to take either a byte array or an existing Database object if you're opening a new connection to the same DB.
diff --git a/src/api.coffee b/src/api.coffee
index f11ea80..311a210 100644
--- a/src/api.coffee
+++ b/src/api.coffee
@@ -237,9 +237,14 @@ class Database
# Open a new database either by creating a new one or opening an existing one,
# stored in the byte array passed in first argument
# @param data [Array<Integer>] An array of bytes representing an SQLite database file
- constructor: (data) ->
- @filename = 'dbfile_' + (0xffffffff*Math.random()>>>0)
- if data? then FS.createDataFile '/', @filename, data, true, true
+ # @param database [Database] An existing database we're opening another connection to
+ constructor: (data, database) ->
+ if database?
+ @filename = database.filename
+ @isconn = true
+ else
+ @filename = 'dbfile_' + (0xffffffff*Math.random()>>>0)
+ if data? then FS.createDataFile '/', @filename, data, true, true
@handleError sqlite3_open @filename, apiTemp
@db = getValue(apiTemp, 'i32')
RegisterExtensionFunctions(@db)
@@ -424,7 +429,7 @@ class Database
removeFunction(func) for _,func of @functions
@functions={}
@handleError sqlite3_close_v2 @db
- FS.unlink '/' + @filename
+ if not @isconn then FS.unlink '/' + @filename
@db = null
### Analyze a result code, return null if no error occured, and throw
+ constructor: (data, database) ->
What if someone wants to open with existing database object but no data object?
Looks nice, though:)
it's a hacky API (mainly so we can use it in a demo tomorrow), but you'd pass undefined as the first param.
I'm not sure what you mean by "whatever transaction isolation between them". Javascript does not have parallel execution, so I don't see in what circumstances having multiple connections would have any advantage.
so I don't see in what circumstances having multiple connections would have any advantage.
It matters because starting transactions often involves opening new connections.
