180816 - Unnecessary nsAutoString construction in HTMLContentSink::CreateContentObject.

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, defect, P2)

Description

[Johnny Stenback (:jst)]

Every time a content object is created we create an nsAutoString that's only
used when creating textarea's, this can be avoided and we'd shave off the over
head of creating (and destroying) the nsAutoString object for ever content
object that's created. Patch coming up.

Comment 1

[Johnny Stenback (:jst)]

Attachment: Don't create an nsAutoString every time we create a content object.

Comment 2

[harishd]

Comment on attachment 106726 [details] [diff] [review]
Don't create an nsAutoString every time we create a content object.

>+    nsCOMPtr<nsIDOMHTMLTextAreaElement> ta(do_QueryInterface(*aResult));
>+    NS_ASSERTION(ta, "Huh? text area doesn't implement "
>+                 "nsIDOMHTMLTextAreaElement?");
>+
>+    ta->SetDefaultValue(Substring(start, end));

You've null check here. no?

r=harishd

Comment 3

[Johnny Stenback (:jst)]

I decided that the null check was not necessary at that point, we'll never
create textarea elements that don't implement nsIDOMHTMLTextAreaElement, if we
start doing that, we'll find out soon enough from the assertion. If someone
disagrees with me I can add back the null check, but IMNSHO the check is
code-bloat we can well live w/o.

Comment 4

[Johnny Stenback (:jst)]

Fixed.

Comment 5

[Alfred Kayser]

Looking at the patch, a question arised:
If the 'skippedContent' is just a CRLF (or CR or LF), 
then it is not empty, but after skipping the initial CRLF, it becomes empty.
Could this happen a lot?, and if so could it save even more by also shaving an
nsAutoString for just a 'CRLF' string?