Fast mseed metadata

collector/FastMSEEDMetadata.py

@@ -0,0 +1,67 @@
+"""Faster implementation of _compute_sample_metrics.
+
+This code reimplements MSEEDMetadata._compute_sample_metrics of
+obspy.signal.quality_control.MSEEDMetadata.
+
+It uses fully vectorised numpy operations which should lead to a significant 
+improvement in speed and memory efficiency.
+
+Author: 
+  @ Roman Racine, 2025
+
+"""
+
+import obspy.signal.quality_control 
+import numpy as np
+
+class FastMSEEDMetadata(obspy.signal.quality_control.MSEEDMetadata):
+    def _compute_sample_metrics(self):
+        """
+        Computes metrics on samples contained in the specified time window
+        """
+        # Make sure there is no integer division by chance.
+        npts = float(self.number_of_samples)
+
+        self.meta['sample_min'] = min([tr.data.min() for tr in self.data])
+        self.meta['sample_max'] = max([tr.data.max() for tr in self.data])
+
+
+        # 
+        # Vectorised implementation to reduce memory consumption and optimise running time
+        #
+        #
+
+        full_samples = np.concatenate([tr.data.astype(np.float64, copy=False) for tr in self.data])
+
+
+        # Manually implement these as they have to work across a list of
+        # arrays.
+        self.meta['sample_mean'] = full_samples.mean()
+
+
+#        full_samples = np.concatenate([tr.data for tr in self.data])
+        self.meta['sample_median'] = np.median(full_samples)
+        self.meta['sample_lower_quartile'] = np.percentile(full_samples, 25)
+        self.meta['sample_upper_quartile'] = np.percentile(full_samples, 75)
+
+
+
+        # more memory efficient implementation of the code below
+        # Roman Racine 2025-09-17
+        self.meta['sample_rms'] = np.sqrt(np.mean(full_samples ** 2))
+
+        # Sample standard deviation
+        self.meta['sample_stdev'] = full_samples.std(ddof=0)
+
+
+        # Percentage based availability as a function of total gap length
+        # over the full trace duration
+        self.meta['percent_availability'] = 100 * (
+            (self.total_time - self.meta['sum_gaps']) /
+            self.total_time)
+
+
+
+
+
+

collector/WFCatalogCollector.py

@@ -87,7 +87,7 @@
 import signal
 import glob
 import re
-
+import FastMSEEDMetadata
 
 def handler(signum, frame):
     raise Exception("Metric calculation has timed out")
@@ -681,15 +681,15 @@ def _callObsPyMetadata(self, files, start, end, granule):
 
                 # Skip continuous segments for hourly granules
                 if granule == "daily":
-                    metadata = MSEEDMetadata(
+                    metadata = FastMSEEDMetadata.FastMSEEDMetadata(
                         files,
                         starttime=start,
                         endtime=end,
                         add_flags=self.args["flags"],
                         add_c_segments=self.args["csegs"],
                     )
                 elif granule == "hourly":
-                    metadata = MSEEDMetadata(
+                    metadata = FastMSEEDMetadata.FastMSEEDMetadata(
                         files,
                         starttime=start,
                         endtime=end,