Perhaps this was premature and I should have created this issue first, but given the simplicity of the change I already submitted a change. If so, that's okay. No hard feelings.

The purpose of MarshalIndent is that one can go from the data structure to the JSON bytes in a single call. You go from 3 calls (NewEncoder, SetIndent, Encode) to just one.

In that sense, EncodeIndent makes less sense. You go from three to two.

In my opinion, if you need a custom encoder, an extra line for the indentation setting doesn't really add complexity. However, adding an extra method to Encoder does add complexity to the package's API.

I always approach both of these outcomes with the same request: "I have this data structure that needs to be marshaled with indention." The only distinguishing factor is "do I already have a Writer?" It is something that I believe is common enough for everyone hence the MarshalIndent function which provides something often used across projects.

However, I find myself needing a direct Encode to an existing io.Writer often enough that warrants a similar function to MarshalIndent that writes to an io.Writer. Perhaps this isn't the correct implementation. Would it be unreasonable for another package-level func that provided an indented marshaling to an io.Writer?

Go generally doesn't optimize for writing one-liners.

As Brad said, it's just one line. It's not worth the complexity of the new API here. If you find yourself writing

enc := json.NewEncoder(os.Stdout)
enc.SetIndent("", "  ")
enc.Encode(&data)

all the time, then instead of writing

json.NewEncoder(os.Stdout).EncodeIndent(&data, "", "  ")

use a custom wrapper that makes that even shorter.