Free temp memory no longer needed in multibyte_split processing

[davidwendt]

Description

Updates the multibyte_split logic to free temporary memory once the chars and offsets have been resolved. This gives room to the remaining processing if more temp memory is required.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[davidwendt]

The change is best viewed with the "Hide whitespace" option. A main section of the source was converted to a IILE which mainly indents many of the lines into a new bracket { } scope.

[bdice]

Looks like most of the diff is code movement/indentation. It looks like the only relevant change is scoping, to allow freeing the temporaries. This looks fine.

[bdice]

I like combining these, it tells me more about the scope of this file (specifically that we won't be putting things into cudf::io or cudf, just cudf::io::text.

Suggested change

	namespace cudf {
	namespace io {
	namespace text {
	namespace cudf::io::text {

[bdice]

The change is best viewed with the "Hide whitespace" option.

I saw this comment just a bit too late! Thanks for the heads-up.

[nvdbaranec]

Only comment I have is that this function is fairly bulky. Might be cleaner to separate this code out into it's own function instead of embedded as a lambda. Minor point though.

[davidwendt]

Only comment I have is that this function is fairly bulky. Might be cleaner to separate this code out into it's own function instead of embedded as a lambda. Minor point though.

Yes, I usually prefer that approach myself but the number of parameters passed in and out seemed like too much here.

[davidwendt]

/merge