feat: better repr for DataType

[NickCrews]

Is your feature request related to a problem?

This is a just a small papercut. If I have a complex dtype, the repr is hard to read, all in one line:

import ibis

dt = ibis.dtype(
    "array<struct<line1: string, line2: string, city: string, state: string, zipcode: string>>"
)
dt
# Array(value_type='Struct([('line1', String(nullable=True)), ('line2', String(nullable=True)), ('city', String(nullable=True)), ('state', String(nullable=True)), ('zipcode', String(nullable=True))], nullable=False), nullable=False)

I think the most readable form would be a pretty-printed version of the string DSL (option A):

dtype("""!array<
    !struct<
        line1: string,
        line2: string,
        city: string,
        state: string,
        zipcode: string
    >
>""")

which is also able to be copy-pasted back into a REPL to recreate the original dtype.

Could also just do the string (option B):

!array<
    !struct<
        line1: string,
        line2: string,
        city: string,
        state: string,
        zipcode: string
    >
>

We should be clever, and only do newlines/pretty if the type is over some metric of complexity, perhaps just string length?

eg I would want array<string> and not

array<
    string
>

I am basing these ideas off of two needs: human readbility, and being able to repro the dtype in the REPL from the output. Are there other considerations I'm not thinking of?

What is the motivation behind your request?

No response

Describe the solution you'd like

see above

What version of ibis are you running?

main

What backend(s) are you using, if any?

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[cpcloud]

We should be clever, and only do newlines/pretty if the type is over some metric of complexity, perhaps just string length?

I'd encourage you to look into the staggering complexity of pretty printing a string before taking any action on this.

We should definitely avoid anything that moves us in the direction of having to consider line breaks and string length when formatting the type string.

[NickCrews]

Ooof, ok thanks for that, I didn't expect it would be that tricky, but I'm sold.

What about going to the more succinct "array<>" DSL, but that's the only change, and keeping everything on one line as it currently is?