| Commit message (Collapse) | Author | Age |
|
|
|
|
|
|
|
|
|
| |
It's better to not add `self` positional argument if it's not needed, instead
of adding it and removing it laster on. `subList` invoaction results in more
allocations and CPU overhead when using positional args.
Closes #5812.
PiperOrigin-RevId: 208687360
|
|
|
|
|
|
|
|
|
|
|
| |
expressions.
This is probably only a theoretical problem, since a blocking struct field is probably a very bad idea.
Closes #5132.
Change-Id: Ie84a78ab4d9ce215f2806ac49bf8911de6959930
PiperOrigin-RevId: 207902766
|
|
|
|
|
| |
RELNOTES: None
PiperOrigin-RevId: 207801155
|
|
|
|
|
|
|
|
| |
This uses SkylarkSemantics now instead of the C++ configuration. The flag is:
--experimental_cc_skylark_api_enabled_packages
RELNOTES:none
PiperOrigin-RevId: 207235431
|
|
|
|
|
|
|
|
|
| |
Java's iterator overhead is significant (2X slower than using indices), so
it's worthwhile to use indices for hot code.
Closes #5737.
PiperOrigin-RevId: 207088309
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
According to JMH using `ImmutableList.of` or `Arrays.asList` is ~2X faster
than using existing approach of creating an empty `ArrayList` with expected
size and populating it using `add` method. This is most likely due to extra
range and capacity checks.
Returning `ImmutableList` instead of `ArrayList` avoids the need to copy it
again in order to create a `SkylarkTuple`.
These changes speed up the function by ~3X.
Closes #5736.
PiperOrigin-RevId: 207079605
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
According to JMH and JIT assembly generated for iterators and index
accesses, iterator methods like `hasNext` and `next` are not optimized
away and result in 3-4X slower execution.
Not using an iterator reduces Buck parse time for FB's internal Android
apps by 7%+.
Closes #5735.
PiperOrigin-RevId: 207073078
|
|
|
|
|
|
|
|
|
| |
Since this happens only every single Skylark method invocation CPU and memory
usage quickly adds up and shows up on profiler samples a lot.
Closes #5656.
PiperOrigin-RevId: 206935520
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This serves 2 purposes:
- better encapsulation and domain model.
- opens the door to optimizations like using `MethodHandle` instead of `Method`
and caching.
- performs one of the optimizations mentioned above - perform `setAccessible`
only once instead of on every method invocation. JMH suggests that this saves
~5% of CPU time.
Next steps are:
- evaluate peformance improvements for some of the optimizations listed above
- make PRs to apply optimizations that seem beneficial.
Closes #5704.
PiperOrigin-RevId: 206805670
|
|
|
|
|
|
|
|
|
|
| |
When `isLegacyNamed` is `true`, `named` is considered to be `true` as well,
so instead of going through an extra indirection and computation, just use
`named` field to store combined result.
Closes #5701.
PiperOrigin-RevId: 206800781
|
|
|
|
|
|
|
|
| |
According to async-profiler, about 50% of `Tuple.getSlice` method invocation is spent in `ImmutableList.copyOf` which is completely unnecessary for cases when an `ImmutableList` instance is passed.
Closes #5699.
PiperOrigin-RevId: 206787490
|
|
|
|
|
|
| |
Closes #5698.
PiperOrigin-RevId: 206780174
|
|
|
|
|
|
|
|
|
|
|
| |
This change is focused on 2 things:
- avoid creating builders in case they don't end up being used
- create builders using the maximum expected size to avoid intermediate
allocations to accommodate more elements
Closes #5694.
PiperOrigin-RevId: 206636046
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Java uses dynamically generated proxy classes to access annotation properties
and their methods are ~7X slower than plain getters. According to async-profiler
50%+ of `convertArgumentList` method time is spent in dynamic proxy methods, so
optimizing their performance makes sense.
This also makes the model less anemic, since POJOs can actually provide business
methods.
Closes #5666.
PiperOrigin-RevId: 206608812
|
|
|
|
|
|
|
|
|
| |
Param annotations only once.
This change has been manually verified to greatly reduce analysis time (~50%) on very large builds, as it mitigates a previous regression brought on by the migration of @SkylarkSignature to @SkylarkCallable.
RELNOTES: None.
PiperOrigin-RevId: 206063684
|
|
|
|
|
|
|
|
|
| |
`keyWordArgs` is already `ImmutableMap<String, Object>` and `copyOf` contains enough
overhead to show up on parse profiles.
Closes #5651.
PiperOrigin-RevId: 206051798
|
|
|
|
|
|
|
| |
Comments are misleading, as discussed on https://github.com/bazelbuild/bazel/pull/5305#issuecomment-396288826
RELNOTES: None.
PiperOrigin-RevId: 205841782
|
|
|
|
|
|
|
|
|
|
| |
`True` instead of `None`.
dict has a get() method that defaults to `None`.
Checking for a key in the dictionary with get() will always return `None` in the given example. Using `True` is better.
RELNOTES: None.
PiperOrigin-RevId: 205052981
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
range used to use MutableList which would eagerly allocate an array list with
all range elements, which is not efficient for very large ranges or when only
a small number of its elements are used.
This implementation uses a constant amount of RAM and computes a value for
each requested index.
For the following Skylark snippet:
```
def check_content(t):
if t == []:
return t
return False
def modulo(n):
return n % 797
N = 10000000
[check_content(i) for i in range(N)]
[check_content(i) for i in range(N)]
[modulo(i) for i in range(N)]
[modulo(i) for i in range(N)]
```
the total runtime goes from
```
$ time bazel-bin/src/main/java/com/google/devtools/skylark/Skylark test.bzl
bazel-bin/src/main/java/com/google/devtools/skylark/Skylark test.bzl 93.09s user 1.67s system 316% cpu 29.930 total
```
to
```
$ time bazel-bin/src/main/java/com/google/devtools/skylark/Skylark test.bzl
bazel-bin/src/main/java/com/google/devtools/skylark/Skylark test.bzl 31.45s user 0.86s system 179% cpu 17.974 total
```
which reflects the reduced system time (fewer allocations) and performance.
Closes #5240.
PiperOrigin-RevId: 204918577
|
|
|
|
|
|
|
| |
parameters of the skylark attr module.
RELNOTES: None.
PiperOrigin-RevId: 204797954
|
|
|
|
|
| |
RELNOTES: None.
PiperOrigin-RevId: 204505003
|
|
|
|
|
|
|
| |
Also add preconditions check.
RELNOTES: None.
PiperOrigin-RevId: 203731929
|
|
|
|
| |
PiperOrigin-RevId: 202690640
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
[Rolling forward https://github.com/bazelbuild/bazel/commit/c4e128e2c6d8cacaeba034d6a3195796d50f1745]
java_common.compile doesn't generate the output source jar when a source jar is
the only input for the compilation. This is wrong because the source jar can
include APT generated sources. It is also inconsistent with java_library and
leads to inconsistent Skylark rules where a declared output will not always have a
generating action.
This new behavior is guarded by a new flag --incompatible_generate_javacommon_source_jar.
RELNOTES: None.
PiperOrigin-RevId: 202648346
|
|
|
|
|
|
|
|
| |
Closes #5307
Closes #5455.
PiperOrigin-RevId: 202567483
|
|
|
|
| |
PiperOrigin-RevId: 202400312
|
|
|
|
|
| |
RELNOTES: None.
PiperOrigin-RevId: 202325860
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Also handle statements in conditional breakpoints.
This is more consistent with other common debuggers (e.g. java, python).
Calls Parser#parseStatement with local parsing level, so some statement types aren't handled (e.g. load statements), which is broadly consistent with other debuggers.
Assignment, augmented assignment, and return statements return a non-None value,
and simple expression statements still return the result of evaluating the expression.
TAG_CHANGE_OK=This proto has never yet been used
TYPE_CHANGE_OK=This proto has never yet been used
PiperOrigin-RevId: 202135678
|
|
|
|
|
| |
RELNOTES: None.
PiperOrigin-RevId: 201748802
|
|
|
|
|
| |
RELNOTES: None
PiperOrigin-RevId: 201617188
|
|
|
|
|
|
|
|
|
| |
instead of object classes.
This allows the caller to, for example, initialize the global library before adding it to a global frame. The skylark interpreter will no longer require that global libraries have a zero-arg constructor.
RELNOTES: None.
PiperOrigin-RevId: 201563608
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
- remove blaze-side thread creation/destruction hooks
- remove ListThreads, ThreadStarted, ThreadEnded events from protocol
- don't track unpaused, not stepping threads in ThreadHandler
The threading model didn't provide useful information -- in practice,
users never want to list currently-running threads, and the debug
client doesn't need to list currently-paused threads, since it receives
each ThreadPaused and ThreadContinued event, so its model is always up to date.
Not tracking thread creation/destruction greatly simplifies the API, and
reduces the blaze-side hooks -- it was also only semi-implemented, as
plenty (/most?) skylark threads were never registered.
The biggest cost to removing this is lack of nice thread names. In
practice, I think the benefits greatly outweigh this cost. If it ever
becomes a problem, there are other lighter-weight ways of communicating
thread names.
TAG_CHANGE_OK=This proto has never yet been used
TYPE_CHANGE_OK=This proto has never yet been used
PiperOrigin-RevId: 201532462
|
|
|
|
|
|
|
|
|
|
| |
useEnvironment
Unfortunately this doesn't work for all callers, namely NativeInfo objects, as they may have structField callables invoked from contexts that have no environment available.
RELNOTES[INC]: Skylark structs (using struct()) may no longer have to_json and to_proto overridden.
PiperOrigin-RevId: 201376969
|
|
|
|
|
|
| |
Get rid of a useless tag, because the Function being tagged is a concrete class, so won't be serializable. Will deal with it in a follow-up. Implement equality for BazelInfo.
PiperOrigin-RevId: 201199255
|
|
|
|
|
| |
RELNOTES: Support for LIPO has been fully removed.
PiperOrigin-RevId: 200724578
|
|
|
|
|
|
|
|
|
| |
Ultimately, we'll need to make the call on whether these functions belong as part of the build API or as part of skylark builtins. For now, we keep them as skylark builtins.
(In either case, we'll want to migrate to @SkylarkCallable, but that's for a later change)
RELNOTES: None.
PiperOrigin-RevId: 200723605
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When pretty printing a Skylark value, lists are presented as the
opening bracket on a line by itself, each entry on its own line,
and the closing bracket again on its own line. While this generally
improves readability, for the empty list this is not the case, as
the expression [] can easily be understood at a glance. In fact, the
additional line even makes the outer structure harder to see, as it
is spread over even more lines. Therefore, shorten the printing of
the empty list to be on a single line.
Change-Id: I032d1550b1f99bce47dbec7e77a4d5c6656d78a1
PiperOrigin-RevId: 200558784
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Environment$Continuation#caller is most definitely nullable in practice.
I suspect it's a bug in skylark itself, but for now, just properly mark
it nullable and handle it in the debugger.
Stack trace from NPE:
Caused by: java.lang.NullPointerException
at com.google.devtools.build.lib.syntax.Environment.listFrames(Environment.java:1197)
at com.google.devtools.build.lib.syntax.Environment.listFrames(Environment.java:81)
at com.google.devtools.build.lib.skylarkdebug.server.ThreadHandler.listFrames(ThreadHandler.java:236)
at com.google.devtools.build.lib.skylarkdebug.server.ThreadHandler.getThreadProto(ThreadHandler.java:345)
at com.google.devtools.build.lib.skylarkdebug.server.ThreadHandler.pauseCurrentThread(ThreadHandler.java:289)
at com.google.devtools.build.lib.skylarkdebug.server.ThreadHandler.pauseIfNecessary(ThreadHandler.java:203)
at com.google.devtools.build.lib.skylarkdebug.server.SkylarkDebugServer.pauseIfNecessary(SkylarkDebugServer.java:158)
at com.google.devtools.build.lib.skylarkdebug.server.SkylarkDebugServer$DebugAwareEval.exec(SkylarkDebugServer.java:262)
at com.google.devtools.build.lib.syntax.UserDefinedFunction.call(UserDefinedFunction.java:91)
at com.google.devtools.build.lib.syntax.BaseFunction.callWithArgArray(BaseFunction.java:462)
at com.google.devtools.build.lib.syntax.BaseFunction.call(BaseFunction.java:440)
at com.google.devtools.build.lib.analysis.skylark.SkylarkRuleConfiguredTargetUtil.lambda$buildRule$1(SkylarkRuleConfiguredTargetUtil.java:105)
at com.google.devtools.build.lib.skylarkdebug.server.SkylarkDebugServer.runWithDebugging(SkylarkDebugServer.java:142)
at com.google.devtools.build.lib.syntax.DebugServerUtils.runWithDebuggingIfEnabled(DebugServerUtils.java:70)
at com.google.devtools.build.lib.analysis.skylark.SkylarkRuleConfiguredTargetUtil.buildRule(SkylarkRuleConfiguredTargetUtil.java:100)
PiperOrigin-RevId: 200229036
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
repository name is remapped.
For example if main/WORKSPACE contains:
local_repository(
name = "a",
path = "../a",
repo_mapping = {"@x" : "@y"},
)
a/BUILD
load("@x//:sample.bzl", "sample")
Then the load in a/BUILD will be resolved as "@y//:sample.bzl"
RELNOTES: None
PiperOrigin-RevId: 200227431
|
|
|
|
|
|
|
|
| |
Switch statements were poorly formatted. Fixing it in a separate commit so
that it doesn't clutter the diff.
RELNOTES: None.
PiperOrigin-RevId: 200062930
|
|
|
|
|
|
| |
- migrate all startTask/completeTask pairs to the new API
PiperOrigin-RevId: 200038703
|
|
|
|
|
|
|
|
|
|
|
|
| |
The high level summary of the changes:
- use `Identifier` instead of `name` in `Keyword` and `Parameter`.
- construct `Identifier` through a factory method in case future interning is desired.
These changes are in preparation for using `Identifier` instead of `name` for environment lookups.
Closes #5304.
PiperOrigin-RevId: 199869171
|
|
|
|
|
|
|
|
|
|
| |
Ensure debugging is turned off (and the server socket closed) when
the command finishes, even if the blaze server is shut down.
Finally, unpause all threads as part of shutting down the debug server,
and also shut down the server if the client connection is lost.
PiperOrigin-RevId: 199863623
|
|
|
|
|
|
|
|
|
|
|
| |
The information was redundant. https://github.com/bazelbuild/bazel/commit/d605a62eb9908fe44c89a2ed10f88f633c4dcb3e modified only one
place, which created a contradiction. We can just remove
the outdated one.
#5352
RELNOTES: None.
PiperOrigin-RevId: 199780557
|
|
|
|
|
|
|
|
|
|
|
|
| |
Not intending this to be a complete implementation yet.
Among the things still to do:
- officially add support for debugging aspects, rules, etc.,
with corresponding tests.
- handle breakpoints at return statements (an edge case which
bypasses Eval).
PiperOrigin-RevId: 199692670
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In particular,
* fix sentences with no space after the period, like [here]([]
* switch backticks to `<code>`, like [here]([]
* link `ctx.var` [here]([]
* Add periods to the end of paragraphs, like [here]([]
* Format a list [here]([]
Closes #5337.
PiperOrigin-RevId: 199631585
|
|
|
|
|
|
|
| |
functionality in WORKSPACE rules.
RELNOTES: None
PiperOrigin-RevId: 199332092
|
|
|
|
|
|
|
|
|
|
| |
There's only one Token and it gets reused.
This reduces the memory usage of the lexer. Parsing time seems to be 5%-10%
faster with this change on a large file. This makes little difference on the
overall performance of Bazel though.
RELNOTES: None.
PiperOrigin-RevId: 199310860
|
|
|
|
|
|
|
| |
Next step will be to skip token allocation.
RELNOTES: None.
PiperOrigin-RevId: 199121625
|
|
|
|
|
|
|
|
|
|
| |
Supporting tools in inputs introduces a slow linear scan. Such tools should be moved to the 'tools' attribute. If --incompatible_no_support_tools_in_action_inputs is set the inputs are scanned, but a helpful error message is issued to the user.
Eventually we will remove the slow scanning. Errors will surface in the execution phase instead of during analysis, and the resulting error messages will be less obvious.
RELNOTES: None
RELNOTES[INC]: With --incompatible_no_support_tools_in_action_inputs enabled, Skylark action inputs are no longer scanned for tools. Move any such inputs to the newly introduced 'tools' attribute.
PiperOrigin-RevId: 198996093
|