Val Town hosts SQLite as part of its standard library (@std/sqlite). This makes a fetch request against their closed-source API (using your API token) which returns results in a consistent format. This is great because you can host your own endpoints that work similarly, and reuse code that was only designed in mind for that original hosted interface
The standard format (abridged to important fields):
{
"statement": {
"sql": "SELECT * FROM some_table",
"args": []
}
}
Output:
{
"rows": [[1, "first", 1709942400], [2, "second", 1709942401]],
"columns": ["id", "name", "lastModified"]
}
{
"statements": [
{
"sql": "INSERT INTO some_table VALUES (?, ?, ?)",
"args": [3, "third", 1709942402]
},
{
"sql": "SELECT * FROM some_table",
"args": []
}
]
}
Output:
[ { "rows": [],
"columns": ["id", "name", "lastModified"]
},
{
"rows": [[1, "first", 1709942400], [2, "second", 1709942401], [3, "third", 1709942402]],
"columns": ["id", "name", "lastModified"]
},
]
There is a deno package (sqlite) which lets you (among other things) create SQLite databases in-memory using WebAssembly. I've created a Val which wraps this to enable it to be a drop-in replacement for @std/sqlite: @postpostscript/sqliteWasm
import { createSqlite } from "https://esm.town/v/postpostscript/sqliteWasm";
import { Statement } from "https://esm.town/v/postpostscript/sqliteBuilder";
const sqlite = createSqlite();
console.log(sqlite.batch([
Statement`
CREATE TABLE test (
id TEXT PRIMARY KEY,
value TEXT
)
`,
Statement`
INSERT INTO test VALUES (
${"some-id"},
${"some-value"}
)
`,
Statement`
SELECT * FROM test
`,
]))
Result:
[ { rows: [],
rowsAffected: 0,
columns: []
},
{
rows: [],
rowsAffected: 1,
columns: []
},
{
rows: [ [ "some-id", "some-value" ] ],
rowsAffected: 0,
columns: [ "id", "value" ]
}
]
I have modified @nbbaier's great work at @postpostscript/sqliteDump to support dumping from any sqlite interface, whether the standard library's version, over HTTP, or through the above Wasm implementation
All of the above enables:
- Serving a subset of your private data publicly for others to query (Example: @postpostscript/sqlitePublic)
- Backing up your database and querying against that backup (via @postpostscript/sqliteBackup's sqliteFromBlob and sqliteToBlob)
But we can do more..! What if we could query from multiple of these data sources.. at the same time! 😱
sqliteUniverse is an @std/sqlite compatible interface that determines where a table should route to based on different patterns
The actual table name will always come after a "/", with the exception of tables without any endpoint, for example users
. Everything before the last "/" is the endpoint name. Endpoint interfaces will be chosen in the following order:
- Exact match in
options.interfaces.exact
e.g.@std/sqlite/someTable
would matchoptions.interfaces.exact["@std/sqlite"]
- Each pattern in
options.interfaces.patterns
options.interfaces.fallback
will be called
An error is thrown if none of the above matches AND returns an sqlite interface. If there is a match but the handler returns nothing, it will continue down the list
patterns.https
-/^https:\/\//
(https://example.com/somePath/tableName
): fetch fromhttps://example.com/somePath/batch
patterns.val
-/^@/
(@author/name/somePath/tableName
): fetch from the val's endpoint,https://author-name.web.val.run/somePath/batch
The following patterns are accessible through import { patterns } from "https://esm.town/v/postpostscript/sqliteUniverse"
:
patterns.blob
-/^blob:\/\//
(blob://backup:sqlite:1709960402936
) - import the database from private blobbackup:sqlite:1709960402936
The sqliteUniverse
export contains defaults insuring no private data will be leaked. If you want to reduce or extend these options, use the sqliteUniverseWithOptions
export and pass a modified interfaces
option in the first argument:
Examples of how to set options.interfaces.exact
, options.interfaces.patterns
, and options.interfaces.fallback
:
import { sqliteUniverseWithOptions, patterns, defaultPatterns } from "https://esm.town/v/postpostscript/sqliteUniverse";
import { createSqlite } from "https://esm.town/v/postpostscript/sqliteWasm";
import { sqliteFromAPI } from "https://esm.town/v/postpostscript/sqliteFromAPI";
import { Statement } from "https://esm.town/v/postpostscript/sqliteBuilder";
const sqlite = sqliteUniverseWithOptions({
interfaces: {
exact: {
// `SELECT * FROM "some-endpoint/someTable"` will match
// `SELECT * FROM "some-endpoint/somePath/someTable"` will NOT match
"some-endpoint": ({ endpoint, tables }) => {
const sqlite = createSqlite()
sqlite.batch([
Statement`CREATE TABLE someTable (someField TEXT PRIMARY KEY)`,
Statement`INSERT INTO someTable VALUES (${"some-field"})`
])
return sqlite
},
},
patterns: [
...defaultPatterns,
[
// shorthand e.g. ~/sqlitePublic -> @postpostscript/sqlitePublic
/^~\/(\w+)/,
({ endpoint, tables, match }) => {
return sqliteFromAPI(`@postpostscript/${match[1]}`)
},
]
],
fallback({ endpoint, tables }) {
// if an endpoint is not found, this will be called
return sqliteFromAPI(`@postpostscript/sqlitePublic`)
},
},
})
console.log(await Statement`
SELECT *
FROM "some-endpoint/someTable"
JOIN "~/sqliteVals/vals"
JOIN authIdExampleComments_comment
LIMIT 1
`.execute({ sqlite }))
Output:
[ { someField: "some-field", id: "aeb70bbb-05fc-403b-8d6a-130c423ecb53", name: "discordWelcomedMembers", code: "// set at Sun Mar 10 2024 00:32:48 GMT+0000 (Coordinated Universal Time)\n" + "export let discordWelcomedM"... 8289 more characters, version: 234771, privacy: "public", public: 1, run_start_at: "2024-03-10T00:32:48.978Z", run_end_at: "2024-03-10T00:32:48.978Z", created_at: "2024-03-10T00:32:48.978Z", author_id: "a0bf3b31-15a5-4d5c-880e-4b1e22c9bc18", author_username: "stevekrouse", username: "postpostscript", comment: "test", date_added: 1709776325.857 }]
Since Val Town currently has a character limit for val readmes, this will have to continue in Part 2!