Consider adding support for string interning during demarshalling

[geeknoid]

It would be nice if the generated demarshalling code could be made to take advantage of a string interning model. This would lead to potentially faster demarshalling, less GC churn, and better processor caching.

What I'm thinking is to have the demarshalling code call out to the GetString function shown below, rather than just calling string([]byte) directly. This would then make it possible to globally install a new interning-savvy function instead.

package string

type GetStringFunc func([]byte) string

var getString GetStringFunc = func(b []byte) string {
	return string(b)
}

func String(b []byte) string {
	return getString(b)
}

func SetFunc(fn GetStringFunc) {
	getString = fn
}

[awalterschulze]

Might be worth looking into this pull request
#217
Maybe this also addresses your concern?

[awalterschulze]

I am closing this because of inactivity, but just comment if you would like it reopened.

[geeknoid]

Sad to close this instead of implementing it :-( Issue #217 does show the inherent value in avoid string excessive string manipulation during demarshalling.

We're about to embark on deep performance work in our project and once all the goofy things are taken care of, I'm sure we'll quickly run into the issue of excessive garbage collection in our code and eliminating allocations will become an important part of improving our perf. Once this happens, I guess we'll send a PR to gogo to implement what this issue is suggesting.

[awalterschulze]

Looking forward to it. I'll reopen.