pydantic v2 support

[sinisaos]

Related to #313.

[codecov-commenter]

Codecov Report

Merging #327 (3295af8) into master (0636086) will not change coverage.
Report is 4 commits behind head on master.
The diff coverage is 100.00%.

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

@@           Coverage Diff           @@
##           master     #327   +/-   ##
=======================================
  Coverage   93.31%   93.31%           
=======================================
  Files           5        5           
  Lines         359      359           
=======================================
  Hits          335      335           
  Misses         24       24           

Files Changed	Coverage Δ
piccolo_admin/endpoints.py	92.61% <100.00%> (ø)

[dantownsend]

Ideally we could do with some kind of function for doing this match. It's unlikely to happen but if the element we need isn't at the first index it would fail.

Something like:

const hasType = (property, type: string): boolean => {
    const types = property.anyOf.map((i) => i.type)
    return types.indexOf(type) != -1
}

[dantownsend]

Hmm, some of them have three potential types:

[dantownsend]

If they're in order, then maybe getting the first element is the way to go.

I wonder what happens if we have a Piccolo column with required=True - does it still return anyof?

[sinisaos]

No. If required=true, not anyOf is returned.

[sinisaos]

Ideally we could do with some kind of function for doing this match. It's unlikely to happen but if the element we need isn't at the first index it would fail.

Something like:

const hasType = (property, type: string): boolean => {
    const types = property.anyOf.map((i) => i.type)
    return types.indexOf(type) != -1
}

In fact, for all columns whose type or format defined in Piccolo ORM pydantic.py, we don't need anyOf. I will make a new commit with those changes. If it's not good, feel free to change it as you think is best.

[dantownsend]

I wonder if it's possible to dramatically simplify this logic. I wonder if there are any column types left now, which when nullable and have an empty string as a value aren't converted to null?

[sinisaos]

Maybe we can use a few variables to reduce code repetition. Something like this

export function convertFormValue(params: {
    key: string
    value: any
    schema: Schema
}): any {
    let { key, value, schema } = params

    const columnTypeAnyOf = schema?.properties[key]["anyOf"][0].type
    const columnFormat = schema?.properties[key].format
    const columnNullable = schema?.properties[key].extra["nullable"] == true
    const possibleFormats = [
        "uuid",
        "email",
        "date-time",
        "date",
        "json",
        "integer",
        "number",
        "string"
    ]

    if (value == "null") {
        value = null
    } else if (schema?.properties[key].type == "array") {
        value = JSON.parse(String(value))
    } else if (possibleFormats.includes(columnFormat) &&
        columnNullable
        && value == "") {
        value = null
    } else if (possibleFormats.includes(columnTypeAnyOf) &&
        columnNullable
        && value == "") {
        value = null
    } else if (schema?.properties[key].extra.foreign_key == true &&
        columnNullable &&
        value == ""
    ) {
        value = null
    }
    return value
}