keystone | Simplifying robust end-to-end machine learning

At the moment there's no way to generate a Keystone schema from an existing Prisma schema. You'd need to create the Keystone schema manually so Keystone could generate a new Prisma schema file.

There's also currently no way to modify the Prisma schema generated by Keystone, though there's been talk of opening this up.

What you want to achieve is called virtual field with the resolver function, you can get all of the item fields and use a string template to construct whatever you need, just put this code as another field of type virtual

record_path is the path to the record, so you should specify the full path

You are trying to destructure a property that doesnt exist on the type.
This should work:

I found the answer eventually.

You have to provide an extra object in your body called variables

Use DropDownSearch() widget for that as below. Add following dependency in your pubspec.yml :

You might be thinking of this the wrong way. Speaking to your specific concerns...

4 separate round trips to the database

Yes, that's true but ideally these would be happening simultaneously. Ie. you shouldn't be waiting for query 1 to return before you fire off query 2. Running the queries in parallel effectively makes the round-trip latency a constant.

How you achieve this comes down to the platform/language your app (the thing issuing these queries) is built in. You many need to reach for some async/await or threading tooling.

4 separate queries that must be made from scratch. The results of some queries should be able to be reused for others. Query 3 < Query 2 < Query 1, and Query 4 < Query 2. That is, the rows of query 2 are a subset of the rows of query 1. So having to go through all the rows again seems bad to me.

It's tempting to think this when you see multiple queries with similar conditions. The (sometimes counter-intuitive) fact is, database are pretty good at resolving queries against a well indexed table and, generally, the simpler the query, the easier it is for the query planner to build a query that will resolve quickly. Trying to be clever by building more complex queries can sometime backfire.

For example, pulling all the records where "championId" = 1 out then performing multiple queries against it might seem like a good idea but, in reality this may require the DB to copy a huge amount of data to memory. Plus, this temporary dataset may not be indexed the same way the original table was, meaning the subsequent queries, although they're running against a smaller dataset, may still be slower.

There definitely are times when isolating a small set of expensive data then querying it multiple ways is the right solution but the example you've given isn't one of them. Put another way, you might be overthinking it. Postgres is a powerful and mature platform that's entire job is consuming queries. Simple comparisons that hit indexed columns will generally be surprisingly fast.

(It all depends on your specifics though so if in doubt, benchmark it.)

Still want to combine multiple queries together even though it's probably not necessary? Ok, let's try it..

A single query is always going to return a single record set. That is, you're always going to get back a "rectangular" set of data, where each row has a value for each column (even if it's null). As such, there's 2 was you can add additional data into a query: by adding more rows or by adding more columns.

First lets setup some test data:

When selecting related items in Keystone, the Admin UI uses item labels – a human readable identifier assigned to each item. The value used for the item label is controlled using the ui.labelField list config:

ui.labelField: Selects the field which will be used as the label column in the Admin UI. By default looks for a field called 'label', then falls back to 'name', then 'title', and finally 'id', which is guaranteed to exist.

– Keystone 6: Schema API docs

In your example, you've configured Keystone to use the firstName field as the label, which works ok, but presents the problem you've identified. What we really need is 3rd field that combines both the first and last names together.

Now, we could do this with hooks, by combining the values and saving them to another text field for example. But this would duplicate the data and we'd need to remember to keep the values in sync, adding complexity. Better to just concatenate these values when they're being read out of the DB. For this, we can use virtual fields.

In the example given, the virtual field solution looks like this:

Have you set a value for the COOKIE_SECRET variable in the Heroku Dashboard? The error you're getting indicates it's undefined or blank.

Heroku calls these "config vars", see the docs.