refactor: Add cached property for long metadata count in POSIX data handling

Author: jay-tau

drishti/handlers/darshan_util.py

@@ -10,6 +10,7 @@
 import pandas as pd
 from darshan import DarshanReport  # type: ignore
 import drishti.includes.parser as parser
+import drishti.includes.config as config
 
 
 class ModuleType(str, Enum):
@@ -272,6 +273,8 @@ class DarshanFile:
     _posix_read_random: Optional[int] = None
     _posix_write_random: Optional[int] = None
 
+    _posix_long_metadata_count: Optional[int] = None
+
     access_pattern: Optional[AccessPatternStats] = None
 
     # Use separate classes for shared operations
@@ -640,4 +643,12 @@ def posix_write_random(self) -> int:
             posix_df = self.report.records[ModuleType.POSIX].to_df()
             posix_counters = posix_df["counters"]
             self._posix_write_random = self.io_stats.get_module_ops(ModuleType.POSIX, "write") - self.posix_write_consecutive - self.posix_write_sequential
-        return self._posix_write_random
+        return self._posix_write_random
+
+    @cached_property
+    def posix_long_metadata_count(self) -> int:
+        if self._posix_long_metadata_count is None:
+            posix_df = self.report.records[ModuleType.POSIX].to_df()
+            posix_long_metadata_rows = posix_df['fcounters'][(posix_df['fcounters']['POSIX_F_META_TIME'] > config.thresholds['metadata_time_rank'][0])]
+            self._posix_long_metadata_count = len(posix_long_metadata_rows)
+        return self._posix_long_metadata_count

drishti/handlers/handle_darshan.py

@@ -530,6 +530,7 @@ def handler():
         shared_files = shared_files.assign(id=lambda d: d['id'].astype(str))
 
         if not shared_files.empty:
+            # TODO: This entire conditional
             total_shared_reads = shared_files['POSIX_READS'].sum()
             total_shared_reads_small = (
                 shared_files['POSIX_SIZE_READ_0_100'].sum() +
@@ -571,7 +572,10 @@ def handler():
 
         count_long_metadata = len(df['fcounters'][(df['fcounters']['POSIX_F_META_TIME'] > thresholds['metadata_time_rank'][0])])
 
-        module.check_long_metadata(count_long_metadata, modules)
+        assert darshan_file_obj.posix_long_metadata_count == count_long_metadata
+        assert darshan_file_obj.modules == modules.keys(), f"{darshan_file_obj.modules} != {modules.keys()}"
+        # module.check_long_metadata(count_long_metadata, modules)
+        module.check_long_metadata(count_long_metadata=darshan_file_obj.posix_long_metadata_count, modules=darshan_file_obj.modules)
 
         # We already have a single line for each shared-file access
         # To check for stragglers, we can check the difference between the