Issue75stopwhendone

[miloyip]

Fix #75
SkipWhitespace() was not modified as @pah suggest, yet.
Actually user may not use this after the stream has parsed a JSON, SkipWhitespace() is only defined for JSON.
Would ask @pah for review.

[pah]

Compared to the discussion in #75, it's a simple but effective approach.

Generally, I liked the idea of making this the default behaviour, except for the insitu parsing.

It seems that it is now possible to read multiple objects without an explicit Reset of the Reader in-between? Should this be required for safety?

SkipWhitespace() was not modified as @pah suggest, yet.
Actually user may not use this after the stream has parsed a JSON, SkipWhitespace() is only defined for JSON.

I don't understand this concern. Yes, it is (kind of) an JSON-related function, but it should work also outside of the Parse* functions?

[miloyip]

Generally, I liked the idea of making this the default behaviour, except for the insitu parsing.

Actually insitu parsing should be fine with kParseStopWhenDoneFlag. These two features should be orthogonal.

It seems that it is now possible to read multiple objects without an explicit Reset of the Reader in-between? Should this be required for safety?

Yes. This is safe for continue parsing with the same stream. Reader does not need Reset() because everytime it passes the stream to ParseStream() but stream of Writer is set at constructor. I forgot about this thus proposed a wrong idea previously.

I don't understand this concern. Yes, it is (kind of) an JSON-related function, but it should work also outside of the Parse* functions?

If the stream contains things other than JSON, I think should not suggest user to use SkipWhitespace() before/after the JSON section. But user can do that if they like to.

[pah]

If the stream contains things other than JSON, I think should not suggest user to use SkipWhitespace() before/after the JSON section. But user can do that if they like to.

Indeed, the user should explictily consume whitespace, if necessary in the application. My suggestion has been to fluently return the stream itself from this function for convenience, e.g. reusing the stream in another context again.

[pah]

Actually insitu parsing should be fine with kParseStopWhenDoneFlag. These two features should be orthogonal.

Currently, there is no way to obtain the "end" of the first JSON object after the insitu parsing, at least for the ParseInsitu(Ch*) overloads in GenericDocument. I think, we should force !kParseStopWhenDone in these cases.

Do you still intend to make this the default behaviour? (i.e. rename it to kParseConsumeFullStream)?

[miloyip]

Insitu parsing can be done with Reader::Parse<kParseInsituFlag>(Stream, Handler) or Document::ParseStream<kParseInsituFlag>(Stream).

For string stream, user also need to call ParseStream() instead of Parse(Ch*), in order to retain the stream between multiple parses.

I am considering setting kParseDefaultFlags = kParseStopWhenDoneFlag.