Use file/offset/kind triplet as elements id.

Use file/offset/kind triplet as elements id.

It saves about 20 MB for each ToDo MVC sample.

R=brianwilkerson@google.com
BUG=

Committed: https://code.google.com/p/dart/source/detail?r=45011

[Brian Wilkerson, 2015-04-09 14:06:25]

LGTM

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/lib/src...
File pkg/analysis_server/lib/src/services/index/store/codec.dart (right):

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/lib/src...
pkg/analysis_server/lib/src/services/index/store/codec.dart:86: LibraryElement
libraryElement = context.getLibraryElement(libSource);
We need to handle the case where 'getLibraryElement' returns 'null'.

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/lib/src...
pkg/analysis_server/lib/src/services/index/store/codec.dart:121: * Returns the
1-st component of the [element] id.
"1-st" --> "first"

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/lib/src...
pkg/analysis_server/lib/src/services/index/store/codec.dart:123: int
encode1(Element element) {
Perhaps more meaningful names, such as "encodeFileName", "encodeOffset", and
"encodeKind" would be better.

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/lib/src...
pkg/analysis_server/lib/src/services/index/store/codec.dart:133: * Returns the
2-nd component of the [element] id.
"2-nd" --> "second"

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/lib/src...
pkg/analysis_server/lib/src/services/index/store/codec.dart:147: * Returns the
3-rd component of the [element] id.
"3-rd" --> "third"

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/lib/src...
File pkg/analysis_server/lib/src/services/index/store/split_store.dart (right):

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/lib/src...
pkg/analysis_server/lib/src/services/index/store/split_store.dart:24: final int
elementId1;
More meaningful names would be good here (and below) as well.

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/test/se...
File pkg/analysis_server/test/services/index/store/codec_test.dart (right):

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/test/se...
pkg/analysis_server/test/services/index/store/codec_test.dart:131: void
test_encode_ConstructorElement_default_synthetic() {
It would be good to have a test case for multiple synthetic constructors (from
mixins).

[scheglov, 2015-04-09 15:11:57]

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/lib/src...
File pkg/analysis_server/lib/src/services/index/store/codec.dart (right):

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/lib/src...
pkg/analysis_server/lib/src/services/index/store/codec.dart:86: LibraryElement
libraryElement = context.getLibraryElement(libSource);
On 2015/04/09 14:06:25, Brian Wilkerson wrote:
> We need to handle the case where 'getLibraryElement' returns 'null'.

Done.

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/lib/src...
pkg/analysis_server/lib/src/services/index/store/codec.dart:121: * Returns the
1-st component of the [element] id.
On 2015/04/09 14:06:25, Brian Wilkerson wrote:
> "1-st" --> "first"

Done.

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/lib/src...
pkg/analysis_server/lib/src/services/index/store/codec.dart:123: int
encode1(Element element) {
On 2015/04/09 14:06:25, Brian Wilkerson wrote:
> Perhaps more meaningful names, such as "encodeFileName", "encodeOffset", and
> "encodeKind" would be better.

The names are intentionally opaque.
Also it's basically an unrolled array.
Clients don't need to know exact meaning of these ids.

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/lib/src...
pkg/analysis_server/lib/src/services/index/store/codec.dart:133: * Returns the
2-nd component of the [element] id.
On 2015/04/09 14:06:25, Brian Wilkerson wrote:
> "2-nd" --> "second"

Done.

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/lib/src...
pkg/analysis_server/lib/src/services/index/store/codec.dart:147: * Returns the
3-rd component of the [element] id.
On 2015/04/09 14:06:25, Brian Wilkerson wrote:
> "3-rd" --> "third"

Done.

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/test/se...
File pkg/analysis_server/test/services/index/store/codec_test.dart (right):

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/test/se...
pkg/analysis_server/test/services/index/store/codec_test.dart:131: void
test_encode_ConstructorElement_default_synthetic() {
On 2015/04/09 14:06:25, Brian Wilkerson wrote:
> It would be good to have a test case for multiple synthetic constructors (from
> mixins).

Done.

[scheglov, 2015-04-09 15:21:23]

Committed patchset #2 (id:20001) manually as 45011 (presubmit successful).

[Brian Wilkerson, 2015-04-09 15:26:31]

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/lib/src...
File pkg/analysis_server/lib/src/services/index/store/codec.dart (right):

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/lib/src...
pkg/analysis_server/lib/src/services/index/store/codec.dart:123: int
encode1(Element element) {
> Clients don't need to know exact meaning of these ids.

I didn't suggest it for clients, I suggested it for developers/maintainers.
Integers just don't provide any semantic information when trying to understand
the code.

[scheglov, 2015-04-09 15:51:08]

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/lib/src...
File pkg/analysis_server/lib/src/services/index/store/codec.dart (right):

https://codereview.chromium.org/1075773002/diff/1/pkg/analysis_server/lib/src...
pkg/analysis_server/lib/src/services/index/store/codec.dart:123: int
encode1(Element element) {
On 2015/04/09 15:26:30, Brian Wilkerson wrote:
> > Clients don't need to know exact meaning of these ids.
> 
> I didn't suggest it for clients, I suggested it for developers/maintainers.
> Integers just don't provide any semantic information when trying to understand
> the code.

I will add comments for these methods.