Automated Publishing

[whittlbc]

Things to get familiar with:

• Cloning a git repo with node.js code | Example - since removed

• Making syscalls in node.js with execSync (such as running a deno file) | Example

• Running a *.ts file with Deno | See ^above example

• How a live object is loaded from a file path | Example

• Extracting a live object's tableSpec | Example

• Converting a live object's tableSpec into the SQL statements needed to actually create it as a Postgres table | Example

• Making raw queries in Typeorm (i.e. SharedTables.query('<sql>', <bindings>)) | Example are all over the code base just grep for them

[AllenOlin]

Steps to tackle #1 of https://github.com/spec-dev/core/blob/master/PUBLISHING.md:

CLI

1. spec publish object <NameOfObject>
2. POST /live-object-version/publish (Lots of logic will be shared from [https://github.com/spec-dev/core/blob/master/core-api/src/routes/admin/live-object-version/liveObjectVersionRoutes.ts#L19])

Payload:
{
    namespace: '<namespace from manifest>',
    name: '<name from manifest>',
    folder: '<input from user>'
}

CORE-API

1. authorize users token for requested nsp (namespace user)
2. kick off delayed job for publishing

DELAYED JOB

Name: publish.ts
1 big job that calls many services inside delayed jobs, unless one of those services needs to be shared.

1. clone github repo from provided link into /tmp example
   • Package.json: "nodegit": "^0.28.0-alpha.18",
2. execSync deno run extractSqlSpec.ts file that:
   a. read and create live object instance from spec.ts in /tmp/ getLiveObjectSpecs
   b. get live object schema/table SQL using: upsertLiveObjectTable
   c. at the bottom of upsertLiveObjectTable, instead of executing a series of psql txs, just return the associated SQL.
   d. pipe SQL return to stdout
3. Do the shitty validation thing with table existence checking
4. use stdout from deno process to do a raw Typeorm query to create schema/tables