Use data driven code when registering constructors.

Use data driven code when registering Element constructors.

For more compact initialization code. Because I want to.

R=adamk@chromium.org

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=167443

[Daniel Bratell, 2014-02-18 18:20:04]

adamk, can you take a look at this? I noticed that these methods shrank a lot
when I improved HashTable::add (and set) which indicated that they had lots of
add/set overhead. Unfortunately that method is more expensive than you would
think so it's better to call in a loop.

No big deal, just a cleanup I did while running past.

[adamk, 2014-02-18 18:28:41]

lgtm once nits are addressed

https://codereview.chromium.org/170793003/diff/1/Source/build/scripts/templat...
File Source/build/scripts/templates/ElementFactory.cpp.tmpl (right):

https://codereview.chromium.org/170793003/diff/1/Source/build/scripts/templat...
Source/build/scripts/templates/ElementFactory.cpp.tmpl:69: const
CreateFunctionMapData data[] = {
Not sure if it matters (since this is only called once), but shouldn't this be
static?

https://codereview.chromium.org/170793003/diff/1/Source/build/scripts/templat...
Source/build/scripts/templates/ElementFactory.cpp.tmpl:74: for (size_t i = 0; i
< sizeof(data) / sizeof(*data); i++)
Nit: you can use WTF_ARRAY_LENGTH() (from wtf/StdLibExtras.h) here instead of
doing sizeof arithmetic.

[Daniel Bratell, 2014-02-19 13:41:05]

https://codereview.chromium.org/170793003/diff/1/Source/build/scripts/templat...
File Source/build/scripts/templates/ElementFactory.cpp.tmpl (right):

https://codereview.chromium.org/170793003/diff/1/Source/build/scripts/templat...
Source/build/scripts/templates/ElementFactory.cpp.tmpl:69: const
CreateFunctionMapData data[] = {
On 2014/02/18 18:28:41, adamk wrote:
> Not sure if it matters (since this is only called once), but shouldn't this be
> static?

It's a tricky question. For a compile time constant "static const" is the right
thing to do. For an array of function and data pointers? Making it static will
create a data block where all the pointers will have to be resolved at a later
time (at worst, when the function is called and then nothing has been gained).

I wasn't sure when I wrote the code what was correct but since I wanted to avoid
the startup cost of resolving pointers I skipped static. Now I've tried with
both (in clang) and it seems static might be better for code size and the extra
cost for loading the binary will be offset by less work when calling this
function.

Short answer: "Yes".

https://codereview.chromium.org/170793003/diff/1/Source/build/scripts/templat...
Source/build/scripts/templates/ElementFactory.cpp.tmpl:74: for (size_t i = 0; i
< sizeof(data) / sizeof(*data); i++)
On 2014/02/18 18:28:41, adamk wrote:
> Nit: you can use WTF_ARRAY_LENGTH() (from wtf/StdLibExtras.h) here instead of
> doing sizeof arithmetic.

Thanks!

[Daniel Bratell, 2014-02-19 13:41:39]

The CQ bit was checked by bratell@opera.com

[commit-bot: I haz the power, 2014-02-19 13:41:56]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/bratell@opera.com/170793003/60001

[Daniel Bratell, 2014-02-19 16:31:10]

The CQ bit was checked by bratell@opera.com

[commit-bot: I haz the power, 2014-02-19 21:58:29]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/bratell@opera.com/170793003/210001

[commit-bot: I haz the power, 2014-02-19 23:03:59]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/bratell@opera.com/170793003/210001

[commit-bot: I haz the power, 2014-02-19 23:47:08]

Change committed as 167443