Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fixed some of the failing parquet_tests [databricks] #11429

Merged
merged 4 commits into from
Sep 10, 2024
Merged

Fixed some of the failing parquet_tests [databricks] #11429

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
d897bff
Fixed some of the failing parquet_tests
razajafri Sep 5, 2024
a865536
Signing off
razajafri Sep 5, 2024
a76e5a5
addressed review comments
razajafri Sep 6, 2024
936203e
removed unused import
razajafri Sep 6, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
47 changes: 26 additions & 21 deletions integration_tests/src/main/python/parquet_test.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,15 +35,19 @@ def read_parquet_df(data_path):
def read_parquet_sql(data_path):
return lambda spark : spark.sql('select * from parquet.`{}`'.format(data_path))

datetimeRebaseModeInWriteKey = 'spark.sql.legacy.parquet.datetimeRebaseModeInWrite' if is_before_spark_400() else 'spark.sql.parquet.datetimeRebaseModeInWrite'
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

All of the non-legacy versions of these configs appear to have been added in 3.0.0. Is there a reason we are not just switching over to using them instead?

int96RebaseModeInWriteKey = 'spark.sql.legacy.parquet.int96RebaseModeInWrite' if is_before_spark_400() else 'spark.sql.parquet.int96RebaseModeInWrite'
datetimeRebaseModeInReadKey = 'spark.sql.legacy.parquet.datetimeRebaseModeInRead' if is_before_spark_400() else 'spark.sql.parquet.datetimeRebaseModeInRead'
int96RebaseModeInReadKey = 'spark.sql.legacy.parquet.int96RebaseModeInRead' if is_before_spark_400() else 'spark.sql.parquet.int96RebaseModeInRead'

rebase_write_corrected_conf = {
'spark.sql.legacy.parquet.datetimeRebaseModeInWrite': 'CORRECTED',
'spark.sql.legacy.parquet.int96RebaseModeInWrite': 'CORRECTED'
datetimeRebaseModeInWriteKey : 'CORRECTED',
int96RebaseModeInWriteKey : 'CORRECTED'
}

rebase_write_legacy_conf = {
'spark.sql.legacy.parquet.datetimeRebaseModeInWrite': 'LEGACY',
'spark.sql.legacy.parquet.int96RebaseModeInWrite': 'LEGACY'
datetimeRebaseModeInWriteKey : 'LEGACY',
int96RebaseModeInWriteKey : 'LEGACY'
}

# Like the standard map_gens_sample but with timestamps limited
Expand Down Expand Up @@ -146,8 +150,8 @@ def test_parquet_read_coalescing_multiple_files(spark_tmp_path, parquet_gens, re
all_confs = copy_and_update(reader_confs, {
'spark.sql.sources.useV1SourceList': v1_enabled_list,
# set the int96 rebase mode values because its LEGACY in databricks which will preclude this op from running on GPU
'spark.sql.legacy.parquet.int96RebaseModeInRead' : 'CORRECTED',
'spark.sql.legacy.parquet.datetimeRebaseModeInRead': 'CORRECTED'})
int96RebaseModeInReadKey : 'CORRECTED',
datetimeRebaseModeInReadKey : 'CORRECTED'})
# once https://github.com/NVIDIA/spark-rapids/issues/1126 is in we can remove spark.sql.legacy.parquet.datetimeRebaseModeInRead config which is a workaround
# for nested timestamp/date support
assert_gpu_and_cpu_are_equal_collect(read_func(data_path),
Expand Down Expand Up @@ -188,8 +192,8 @@ def test_parquet_read_round_trip(spark_tmp_path, parquet_gens, read_func, reader
all_confs = copy_and_update(reader_confs, {
'spark.sql.sources.useV1SourceList': v1_enabled_list,
# set the int96 rebase mode values because its LEGACY in databricks which will preclude this op from running on GPU
'spark.sql.legacy.parquet.int96RebaseModeInRead' : 'CORRECTED',
'spark.sql.legacy.parquet.datetimeRebaseModeInRead': 'CORRECTED'})
int96RebaseModeInReadKey : 'CORRECTED',
datetimeRebaseModeInReadKey : 'CORRECTED'})
# once https://github.com/NVIDIA/spark-rapids/issues/1126 is in we can remove spark.sql.legacy.parquet.datetimeRebaseModeInRead config which is a workaround
# for nested timestamp/date support
assert_gpu_and_cpu_are_equal_collect(read_func(data_path),
Expand All @@ -199,6 +203,7 @@ def test_parquet_read_round_trip(spark_tmp_path, parquet_gens, read_func, reader
@allow_non_gpu('FileSourceScanExec')
@pytest.mark.parametrize('read_func', [read_parquet_df, read_parquet_sql])
@pytest.mark.parametrize('disable_conf', ['spark.rapids.sql.format.parquet.enabled', 'spark.rapids.sql.format.parquet.read.enabled'])
@disable_ansi_mode
def test_parquet_fallback(spark_tmp_path, read_func, disable_conf):
data_gens = [string_gen,
byte_gen, short_gen, int_gen, long_gen, boolean_gen] + decimal_gens
Expand All @@ -225,8 +230,8 @@ def test_parquet_read_round_trip_binary(std_input_path, read_func, binary_as_str
all_confs = copy_and_update(reader_confs, {
'spark.sql.parquet.binaryAsString': binary_as_string,
# set the int96 rebase mode values because its LEGACY in databricks which will preclude this op from running on GPU
'spark.sql.legacy.parquet.int96RebaseModeInRead' : 'CORRECTED',
'spark.sql.legacy.parquet.datetimeRebaseModeInRead': 'CORRECTED'})
int96RebaseModeInReadKey : 'CORRECTED',
datetimeRebaseModeInReadKey : 'CORRECTED'})
# once https://github.com/NVIDIA/spark-rapids/issues/1126 is in we can remove spark.sql.legacy.parquet.datetimeRebaseModeInRead config which is a workaround
# for nested timestamp/date support
assert_gpu_and_cpu_are_equal_collect(read_func(data_path),
Expand All @@ -245,8 +250,8 @@ def test_binary_df_read(spark_tmp_path, binary_as_string, read_func, data_gen):
all_confs = {
'spark.sql.parquet.binaryAsString': binary_as_string,
# set the int96 rebase mode values because its LEGACY in databricks which will preclude this op from running on GPU
'spark.sql.legacy.parquet.int96RebaseModeInRead': 'CORRECTED',
'spark.sql.legacy.parquet.datetimeRebaseModeInRead': 'CORRECTED'}
int96RebaseModeInReadKey : 'CORRECTED',
datetimeRebaseModeInReadKey : 'CORRECTED'}
assert_gpu_and_cpu_are_equal_collect(read_func(data_path), conf=all_confs)

@pytest.mark.parametrize('v1_enabled_list', ["", "parquet"])
Expand All @@ -256,8 +261,8 @@ def test_parquet_read_forced_binary_schema(std_input_path, v1_enabled_list):
all_confs = copy_and_update(reader_opt_confs[0], {
'spark.sql.sources.useV1SourceList': v1_enabled_list,
# set the int96 rebase mode values because its LEGACY in databricks which will preclude this op from running on GPU
'spark.sql.legacy.parquet.int96RebaseModeInRead' : 'CORRECTED',
'spark.sql.legacy.parquet.datetimeRebaseModeInRead': 'CORRECTED'})
int96RebaseModeInReadKey : 'CORRECTED',
datetimeRebaseModeInReadKey : 'CORRECTED'})
# once https://github.com/NVIDIA/spark-rapids/issues/1126 is in we can remove spark.sql.legacy.parquet.datetimeRebaseModeInRead config which is a workaround
# for nested timestamp/date support

Expand All @@ -277,8 +282,8 @@ def test_parquet_read_round_trip_binary_as_string(std_input_path, read_func, rea
'spark.sql.sources.useV1SourceList': v1_enabled_list,
'spark.sql.parquet.binaryAsString': 'true',
# set the int96 rebase mode values because its LEGACY in databricks which will preclude this op from running on GPU
'spark.sql.legacy.parquet.int96RebaseModeInRead' : 'CORRECTED',
'spark.sql.legacy.parquet.datetimeRebaseModeInRead': 'CORRECTED'})
int96RebaseModeInReadKey : 'CORRECTED',
datetimeRebaseModeInReadKey : 'CORRECTED'})
# once https://github.com/NVIDIA/spark-rapids/issues/1126 is in we can remove spark.sql.legacy.parquet.datetimeRebaseModeInRead config which is a workaround
# for nested timestamp/date support
assert_gpu_and_cpu_are_equal_collect(read_func(data_path),
Expand Down Expand Up @@ -342,16 +347,16 @@ def test_parquet_read_roundtrip_datetime_with_legacy_rebase(spark_tmp_path, parq
gen_list = [('_c' + str(i), gen) for i, gen in enumerate(parquet_gens)]
data_path = spark_tmp_path + '/PARQUET_DATA'
write_confs = {'spark.sql.parquet.outputTimestampType': ts_type,
'spark.sql.legacy.parquet.datetimeRebaseModeInWrite': ts_rebase_write[0],
'spark.sql.legacy.parquet.int96RebaseModeInWrite': ts_rebase_write[1]}
datetimeRebaseModeInWriteKey : ts_rebase_write[0],
int96RebaseModeInWriteKey : ts_rebase_write[1]}
with_cpu_session(
lambda spark: gen_df(spark, gen_list).write.parquet(data_path),
conf=write_confs)
# The rebase modes in read configs should be ignored and overridden by the same modes in write
# configs, which are retrieved from the written files.
read_confs = copy_and_update(reader_confs, {'spark.sql.sources.useV1SourceList': v1_enabled_list,
'spark.sql.legacy.parquet.datetimeRebaseModeInRead': ts_rebase_read[0],
'spark.sql.legacy.parquet.int96RebaseModeInRead': ts_rebase_read[1]})
datetimeRebaseModeInReadKey : ts_rebase_read[0],
int96RebaseModeInReadKey : ts_rebase_read[1]})
assert_gpu_and_cpu_are_equal_collect(
lambda spark: spark.read.parquet(data_path),
conf=read_confs)
Expand Down Expand Up @@ -734,7 +739,7 @@ def test_nested_pruning_and_case_insensitive(spark_tmp_path, data_gen, read_sche
all_confs = copy_and_update(reader_confs, {
'spark.sql.sources.useV1SourceList': v1_enabled_list,
'spark.sql.optimizer.nestedSchemaPruning.enabled': nested_enabled,
'spark.sql.legacy.parquet.datetimeRebaseModeInRead': 'CORRECTED'})
datetimeRebaseModeInReadKey : 'CORRECTED'})
# This is a hack to get the type in a slightly less verbose way
rs = StructGen(read_schema, nullable=False).data_type
assert_gpu_and_cpu_are_equal_collect(lambda spark : spark.read.schema(rs).parquet(data_path),
Expand Down
5 changes: 3 additions & 2 deletions sql-plugin/src/main/spark400/scala/com/nvidia/spark/rapids/shims/GpuBatchScanExec.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,7 +32,7 @@ import org.apache.spark.sql.catalyst.util.{truncatedString, InternalRowComparabl
import org.apache.spark.sql.connector.catalog.Table
import org.apache.spark.sql.connector.read._
import org.apache.spark.sql.execution.datasources.rapids.DataSourceStrategyUtils
import org.apache.spark.sql.execution.datasources.v2.{DataSourceRDD, StoragePartitionJoinParams}
import org.apache.spark.sql.execution.datasources.v2._
import org.apache.spark.sql.internal.SQLConf

case class GpuBatchScanExec(
Expand All @@ -47,7 +47,7 @@ case class GpuBatchScanExec(
@transient override lazy val batch: Batch = if (scan == null) null else scan.toBatch
// TODO: unify the equal/hashCode implementation for all data source v2 query plans.
override def equals(other: Any): Boolean = other match {
case other: GpuBatchScanExec =>
case other: BatchScanExec =>
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why is this being changed? This is a GpuBatchScanExec. We don't want to be equal to non-GPU versions do we?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Right. While debugging I wasn't sure what was causing a failure and looking at 330 shim I changed this and didn't change it back before submitting this PR.

I am adding that change to this PR as well.

this.batch != null && this.batch == other.batch &&
this.runtimeFilters == other.runtimeFilters &&
this.spjParams == other.spjParams
Expand Down Expand Up @@ -137,6 +137,7 @@ case class GpuBatchScanExec(
override lazy val readerFactory: PartitionReaderFactory = batch.createReaderFactory()

override lazy val inputRDD: RDD[InternalRow] = {
scan.metrics = allMetrics
val rdd = if (filteredPartitions.isEmpty && outputPartitioning == SinglePartition) {
// return an empty RDD with 1 partition if dynamic filtering removed the only split
sparkContext.parallelize(Array.empty[InternalRow], 1)
Expand Down