Enable PLC1802 on ruff (#165813)

This PR enables ruff check `PLC1802`, which detects len calls on sequences in a boolean test context.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/165813
Approved by: https://github.com/ezyang

benchmarks/dynamo/huggingface.py

@@ -124,7 +124,7 @@ with open(MODELS_FILENAME) as fh:
             continue
         batch_size = int(batch_size)
         BATCH_SIZE_KNOWN_MODELS[model_name] = batch_size
-assert len(BATCH_SIZE_KNOWN_MODELS)
+assert BATCH_SIZE_KNOWN_MODELS
 
 
 try:

pyproject.toml

@@ -212,6 +212,7 @@ select = [
     "PIE810",
     "PLC0131", # type bivariance
     "PLC0132", # type param mismatch
+    "PLC1802", # len({expression}) used as condition without comparison
     "PLC0205", # string as __slots__
     "PLC3002", # unnecessary-direct-lambda-call
     "PLE",

test/quantization/core/test_quantized_tensor.py

@@ -100,7 +100,7 @@ def param_search_greedy(x, bit_rate, n_bins=200, ratio=0.16):
             cur_min, cur_max, cur_loss = cur_min + stepsize, cur_max, loss1
         else:
             cur_min, cur_max, cur_loss = cur_min, cur_max - stepsize, loss2
-    if len(solutions):
+    if solutions:
         best = solutions[0]
         for solution in solutions:
             if solution[-1] < best[-1]:

torch/_dynamo/backends/distributed.py

@@ -98,14 +98,14 @@ def pretty_print_buckets(buckets: list[Bucket], bucket_bytes_cap: int) -> None:
                 )
             )
 
-    if len(rows):
+    if rows:
         log.info(
             "\nDDPOptimizer used bucket cap %s and created %d buckets. Enable debug logs for detailed bucket info.",
             bucket_bytes_cap,
             len(buckets),
         )
 
-        if len(extended_buckets):
+        if extended_buckets:
             log.warning(
                 "Some buckets were extended beyond their requested parameter capacities"
                 " in order to ensure each subgraph has an output node, required for fx graph partitioning."
@@ -122,7 +122,7 @@ def pretty_print_buckets(buckets: list[Bucket], bucket_bytes_cap: int) -> None:
                 tabulate(rows, headers=headers, tablefmt="simple_grid"),
             )
 
-            if len(extended_buckets):
+            if extended_buckets:
                 log.warning(
                     "DDPOptimizer extended these buckets to ensure per-subgraph output nodes:\n%s",
                     tabulate(

torch/_dynamo/output_graph.py

@@ -1867,7 +1867,7 @@ class OutputGraph(OutputGraphCommon):
                 _get_source_debug_name(var.source) for var in potential_side_effects
             ]
 
-            if len(side_effect_refs):
+            if side_effect_refs:
                 warnings.warn(
                     f"While exporting, we found certain side effects happened in the model.forward. "
                     f"Here are the list of potential sources you can double check: {side_effect_refs}"
@@ -3736,7 +3736,7 @@ class SubgraphTracer(fx.Tracer):
             if v1 != v2
         ]
 
-        if len(mutated_inputs):
+        if mutated_inputs:
             mutated_nodes = [input_nodes[i] for i in mutated_inputs]
             msg = f"Input mutation detected at {mutated_nodes}"
             return MutationInfo(True, msg)

torch/_dynamo/variables/builtin.py

@@ -1847,7 +1847,7 @@ class BuiltinVariable(VariableTracker):
                 polyfills.builtins.iter_
             ).call_function(tx, [obj, *args], {})
 
-            if len(args):
+            if args:
                 # iter(obj, sentinel) returns an object that implements
                 # __iter__ and __next__ methods (UserDefinedObjectVariable)
                 # Wrap the return value in a IteratorVariable subclass (LazyObjectIteratorVariable)

torch/_inductor/comms.py

@@ -834,7 +834,7 @@ def _schedule_for_comm(
             collective_cost -= snode_to_cost[candidate.snode]
         heapq.heapify(ready)
 
-    while len(ready):
+    while ready:
         snode = heapq.heappop(ready).snode
         if reorder_for_overlap and contains_collective(snode):
             schedule_collective_for_overlap(snode)

torch/_inductor/runtime/triton_heuristics.py

@@ -2895,7 +2895,7 @@ def match_target_block_product(
         relative_scores[dim] = score / total_score
 
     # Scale up dimensions by their relative scores until we reach the target
-    while curr_block_product < target_block_product and len(relative_scores):
+    while curr_block_product < target_block_product and relative_scores:
         dim, score = max(relative_scores.items(), key=lambda item: item[1])
 
         # Check if we've hit the max for this dimension

torch/_inductor/utils.py

@@ -792,7 +792,7 @@ def get_kernel_metadata(
     # where `inductor_nodes` contains nodes from multiple graph instances
     # is not supported. An example of this is conditional statements.
     single_graph = None
-    if len(inductor_nodes):
+    if inductor_nodes:
         unique_graphs = OrderedSet(n.graph for n in inductor_nodes)
         if len(unique_graphs) == 1:
             single_graph = inductor_nodes[0].graph

torch/distributed/fsdp/_fully_shard/_fsdp_param_group.py

@@ -237,7 +237,7 @@ class FSDPParamGroup:
             raise AssertionError(
                 f"FSDP expects uniform original parameter dtype but got {orig_dtypes}"
             )
-        self._orig_dtype = next(iter(orig_dtypes)) if len(trainable_params) else None
+        self._orig_dtype = next(iter(orig_dtypes)) if trainable_params else None
         if len(trainable_params) > 0 and len(reduce_dtypes) != 1:
             # This can be relaxed if we issue one reduce-scatter per reduce
             # dtype (but we would need a way for users to specify multiple
@@ -245,9 +245,7 @@ class FSDPParamGroup:
             raise AssertionError(
                 f"FSDP expects uniform reduce dtype but got {reduce_dtypes}"
             )
-        self._reduce_dtype = (
-            next(iter(reduce_dtypes)) if len(trainable_params) else None
-        )
+        self._reduce_dtype = next(iter(reduce_dtypes)) if trainable_params else None
 
     def lazy_init(self):
         # Lazy init should be idempotent

torch/distributed/pipelining/schedules.py

@@ -2178,7 +2178,7 @@ BACKWARD_INPUT, BACKWARD_WEIGHT, and OVERLAP_F_B are supported."
                 raise e
 
         # Mostly these operations should have finished long ago, but there isn't an obvious time when to wait for them
-        while len(send_ops):
+        while send_ops:
             _wait_batch_p2p(send_ops.pop())
 
         assert len(self.unshard_ops) == 0, "Unused unshard operations"

torch/hub.py

@@ -372,7 +372,7 @@ def _check_dependencies(m):
 
     if dependencies is not None:
         missing_deps = [pkg for pkg in dependencies if not _check_module_exists(pkg)]
-        if len(missing_deps):
+        if missing_deps:
             raise RuntimeError(f"Missing dependencies: {', '.join(missing_deps)}")
 
 

torch/testing/_internal/opinfo/core.py

@@ -166,7 +166,7 @@ class SampleInput:
 A SampleInput can be constructed "naturally" with *args and **kwargs or by
 explicitly setting the "args" and "kwargs" parameters, but the two
 methods of construction cannot be mixed!"""
-        elif len(var_args) or len(var_kwargs):
+        elif var_args or var_kwargs:
             assert (
                 output_process_fn_grad is None
                 and broadcasts_input is None

torch/utils/data/datapipes/dataframe/datapipes.py

@@ -53,7 +53,7 @@ class ConcatDataFramesPipe(DFIterDataPipe):
             if len(buffer) == self.n_batch:
                 yield df_wrapper.concat(buffer)
                 buffer = []
-        if len(buffer):
+        if buffer:
             yield df_wrapper.concat(buffer)
 
 
@@ -78,7 +78,7 @@ class ShuffleDataFramesPipe(DFIterDataPipe):
             if len(buffer) == size:
                 yield df_wrapper.concat(buffer)
                 buffer = []
-        if len(buffer):
+        if buffer:
             yield df_wrapper.concat(buffer)
 
 
@@ -107,7 +107,7 @@ class FilterDataFramesPipe(DFIterDataPipe):
                 if len(buffer) == size:
                     yield df_wrapper.concat(buffer)
                     buffer = []
-        if len(buffer):
+        if buffer:
             yield df_wrapper.concat(buffer)
 
 

torch/utils/data/datapipes/iter/combining.py

@@ -626,7 +626,7 @@ class MultiplexerIterDataPipe(IterDataPipe):
 
     def __iter__(self):
         iterators = [iter(x) for x in self.datapipes]
-        while len(iterators):
+        while iterators:
             for it in iterators:
                 try:
                     value = next(it)

torch/utils/data/datapipes/iter/selecting.py

@@ -88,7 +88,7 @@ class FilterIterDataPipe(IterDataPipe[_T_co]):
             for idx, mask in enumerate(df_wrapper.iterate(condition)):
                 if mask:
                     result.append(df_wrapper.get_item(data, idx))
-            if len(result):
+            if result:
                 return True, df_wrapper.concat(result)
             else:
                 return False, None  # type: ignore[return-value]

torch/utils/weak.py

@@ -309,7 +309,7 @@ class WeakIdKeyDictionary(MutableMapping):
                 dict = type({})(dict)
             for key, value in dict.items():
                 d[self.ref_type(key, self._remove)] = value  # CHANGED
-        if len(kwargs):
+        if kwargs:
             self.update(kwargs)
 
     def __ior__(self, other):