I'm writing a backend that works with the

package

goal to package 4 distinct target types. To do this, I need to implement a

PackageFieldSet

. Essentially, I need all the fields on the target, which means I need a

FieldSet

that lists all of the fields on the target. I'm debating between making a

FieldSet

for each target that lists everything, or creating a single

FieldSet

that lists the few fields that are common between the 4 distinct target types. But, I'm not sure what the value of having a

FieldSet

is. From the docs:

A

FieldSet

is a way to specify which Fields your rule needs to use in a typed way that is understood by the engine.

Normally, your rule should simply use

tgt.get()

and

tgt.has_field()

instead of a

FieldSet

. However, for several of the Common plugin tasks, you will instead need to create a

FieldSet

so that the combination of fields you use can be represented by a type understood by the engine.

https://www.pantsbuild.org/v2.17/docs/rules-api-and-target-api#fieldsets What is the value of listing all of the fields in the

FieldSet

? Could I list a bare minimum set in the

FieldSet

, get the original target, and then access the fields on the original target? Why does the engine need the types on the

FieldSet

for packaging? Something to do with caching?

At a broad brush, a target in a build file will generally collect more metadata than most goals / rules need; so the field set lets a rule say exactly what it needs - for caching granularity purposes as you guessed.

Ah. I can see the distinction for targets with sources. That makes sense.