From 2ec12ddb027225ab45affa494ec739c417f44727 Mon Sep 17 00:00:00 2001
From: Derek Feichtinger <derek.feichtinger@psi.ch>
Date: Mon, 6 Jul 2026 15:12:45 +0200
Subject: [PATCH] implement generic way of storing metric vectors in
 aggregation

Already move sdev to use the new field instead of specialized
hardcoded fields.
---
 slurm-eff-tool.py | 29 ++++++++++++++++-------------
 1 file changed, 16 insertions(+), 13 deletions(-)

diff --git a/slurm-eff-tool.py b/slurm-eff-tool.py
index 8957af0..727a609 100755
--- a/slurm-eff-tool.py
+++ b/slurm-eff-tool.py
@@ -482,9 +482,7 @@ class OutputRow:
     waste_CPU: float | None   # CPU h
     waste_Mem: float | None
     waste_total: float | None
-    _cpu_eff_values: list[float] = field(default_factory=list, repr=False)
-    _mem_eff_values: list[float] = field(default_factory=list, repr=False)
-    _time_eff_values: list[float] = field(default_factory=list, repr=False)
+    vectors: dict[str, list[int | float]] # here we can store full vectors of the aggreg.
 
     def as_dict(self, sdev: bool = False) -> dict[str, Any]:
         d: dict[str, Any] = {
@@ -516,9 +514,9 @@ class OutputRow:
         }
         if sdev:
             for col, vals in [
-                ("CPU_Eff", self._cpu_eff_values),
-                ("Mem_Eff", self._mem_eff_values),
-                ("Time_Eff", self._time_eff_values),
+                ("CPU_Eff", self.vectors['cpu_eff']),
+                ("Mem_Eff", self.vectors['mem_eff']),
+                ("Time_Eff", self.vectors['time_eff']),
             ]:
                 d[f"{col}_sdev"] = stdev_or_none(vals)
                 d[f"{col}_max"] = max(vals) if vals else None
@@ -1016,9 +1014,9 @@ def make_single_row(rec: JobRecord, dflt_mpcpu: float) -> OutputRow:
         waste_Mem=waste_mem,
         waste_CPU=waste_cpu,
         waste_total=waste_total,
-        _cpu_eff_values=[rec.cpu_eff] if rec.cpu_eff is not None else [],
-        _mem_eff_values=[rec.mem_eff] if rec.mem_eff is not None else [],
-        _time_eff_values=[rec.time_eff] if rec.time_eff is not None else [],
+        vectors = {"cpu_eff": [rec.cpu_eff] if rec.cpu_eff is not None else [],
+                   "mem_eff": [rec.mem_eff] if rec.mem_eff is not None else [],
+                   "time_eff": [rec.time_eff] if rec.time_eff is not None else []},
     )
 
 
@@ -1073,7 +1071,14 @@ def make_aggregate_row(records: list[JobRecord], username: str, partition: str,
             waste_cpu = count * walltime * (100-cpu_efficiency)/100 * first.cpus
 
     waste_total = waste_cpu + waste_mem/dflt_mpcpu
-    
+
+    vectors: dict[str, list[int | float]] = {
+        "cpu_eff": [r.cpu_eff for r in records if r.cpu_eff is not None],
+        "mem_eff": [r.mem_eff for r in records if r.mem_eff is not None],
+        "time_eff": [r.time_eff for r in records if r.time_eff is not None],
+    }
+
+
     return OutputRow(
         username=username,
         JobID="",
@@ -1100,9 +1105,7 @@ def make_aggregate_row(records: list[JobRecord], username: str, partition: str,
         waste_Mem=waste_mem,
         waste_CPU=waste_cpu,
         waste_total=waste_total,
-        _cpu_eff_values=cpu_eff_vals,
-        _mem_eff_values=mem_eff_vals,
-        _time_eff_values=time_eff_vals,
+        vectors = vectors,
     )
 
 def resolve_column_name(name: str) -> str: