add dataref export

parse_results.py

@@ -6,6 +6,7 @@ import csv
 import fnmatch
 import io
 import re
+import sys
 from pathlib import Path
 from typing import Dict, Iterable, List, Mapping, Optional, Sequence, Union
 
@@ -100,7 +101,7 @@ def collect_results(result_dir: Union[Path, str],
                     exp_data[k] = to_number(configparser['global'][k])
                 except KeyError:
                     error = True
-                    print(f'ERROR in {data_path}')
+                    print(f'ERROR in {data_path}', file=sys.stderr)
                     break
 
             if error:
@@ -182,10 +183,11 @@ def calculate_stats(data: EvaluationResults, warn=False) -> EvaluationStats:
                 if warn:
                     precision = cur_stats['std'] / cur_stats['mean']
                     if precision > 0.05:
-                        print('Warning: imprecise data impl:', end='')
-                        print(
+                        print((
+                            'Warning: imprecise data impl:'
                             f' {impl}, c: {cons}, k: {key} p: {precision*100:.5}%'
-                        )
+                        ),
+                              file=sys.stderr)
 
                 cons_stats[key] = cur_stats
 
@@ -218,6 +220,23 @@ def print_stats(stats: EvaluationStats, variables: Optional[Iterable]):
                 print_stats_for_variable(variable, cons_stats[variable])
 
 
+def print_dataref(stats: EvaluationStats, variables=None, dstats=None):
+    for impl, impl_stats in stats.items():
+        for connections, conn_stats in impl_stats.items():
+            keys = variables or conn_stats.keys()
+            for key in keys:
+                desc_stats = conn_stats[key]
+
+                dstats = dstats or desc_stats.keys()
+                for stat in dstats:
+                    val = desc_stats[stat]
+                    if stat == 'outliers':
+                        continue
+                    print(
+                        rf'\drefset{{{impl}/{connections}/{key}/{stat}}}{{{val}}}'
+                    )
+
+
 def main():
     parser = argparse.ArgumentParser()
     parser.add_argument("result_dir")
@@ -242,6 +261,14 @@ def main():
         nargs='*',
         default=['iops'])
 
+    parser.add_argument("--dataref",
+                        help="dump stats as dataref",
+                        action='store_true')
+
+    parser.add_argument("--desc-stats",
+                        help="descriptive statistics to print",
+                        nargs='*')
+
     args = parser.parse_args()
 
     data = collect_results(args.result_dir,
@@ -251,6 +278,12 @@ def main():
 
     stats = calculate_stats(data, args.warn)
 
+    if args.dataref:
+        print_dataref(stats,
+                      variables=args.print_stats,
+                      dstats=args.desc_stats)
+        return
+
     if args.print_stats is not None:
         print_stats(stats, variables=args.print_stats)
 

plot_tail_latency.py

@@ -179,7 +179,7 @@ def plot_tail_latency_linear(latency_files, latency, out=None):
     pyplot_show_or_save(out=out)
 
 
-def summarize(latency_files, latency):
+def summarize(latency_files, latency, dataref=False):
     bar_data = get_bar_data(latency_files,
                             latency=latency,
                             include_raw_data=True)
@@ -197,28 +197,50 @@ def summarize(latency_files, latency):
             data[99.99],
         ]) / 1000000
 
-        name = impl.replace('-io', '').replace('emper-',
-                                               '').replace('-0us-1ms', '')
+        name = impl.replace('-io', '').replace('emper-', '')
+
         data_min = min(latencies)
         data_max = max(latencies)
-        print(f'{name}-min: {data_min:.2f}')
-        print(f'{name}-max: {data_max:.2f}')
-        print(f'{name}-mean: {np.mean(latencies):.2f}')
-
-        print(f'{name}-median: {median:.2f}')
-        print(f'{name}-p95: {p95:.2f}')
-        print(f'{name}-p99: {p99:.2f}')
-        print(f'{name}-p99.9: {p99_9:.2f}')
-        print(f'{name}-p99.99: {p99_99:.2f}')
-
-        bin_edge_candidates = [
-            data_min, median, 100, 500, 1000, 1750, 2500, 5000, 7500, 10000,
-            13000, 13500, 14000, data_max
-        ]
-        bin_edges = [c for c in bin_edge_candidates if c <= data_max]
-        hist, bin_edges = np.histogram(latencies, bin_edges)
-        print(f'{name}-hist: {list(hist)}')
-        print(f'{name}-hist-bin-edges: {[f"{e:.2f}" for e in bin_edges]}')
+
+        if not dataref:
+            print(f'{name}-min: {data_min:.2f}')
+            print(f'{name}-max: {data_max:.2f}')
+            print(f'{name}-mean: {np.mean(latencies):.2f}')
+
+            print(f'{name}-median: {median:.2f}')
+            print(f'{name}-p95: {p95:.2f}')
+            print(f'{name}-p99: {p99:.2f}')
+            print(f'{name}-p99.9: {p99_9:.2f}')
+            print(f'{name}-p99.99: {p99_99:.2f}')
+
+            bin_edge_candidates = [
+                data_min, median, 100, 500, 1000, 1750, 2500, 5000, 7500,
+                10000, 13000, 13500, 14000, data_max
+            ]
+            bin_edges = [c for c in bin_edge_candidates if c <= data_max]
+            hist, bin_edges = np.histogram(latencies, bin_edges)
+            print(f'{name}-hist: {list(hist)}')
+            print(f'{name}-hist-bin-edges: {[f"{e:.2f}" for e in bin_edges]}')
+        else:
+            print(
+                rf'\drefset[unit=\unit{{\ms}}]{{{name}/latency/min}}{{{data_min}}}'
+            )
+            print(
+                rf'\drefset[unit=\unit{{\ms}}]{{{name}/latency/max}}{{{data_max}}}'
+            )
+            print(
+                rf'\drefset[unit=\unit{{\ms}}]{{{name}/latency/median}}{{{median}}}'
+            )
+            print(
+                rf'\drefset[unit=\unit{{\ms}}]{{{name}/latency/p95}}{{{p95}}}')
+            print(
+                rf'\drefset[unit=\unit{{\ms}}]{{{name}/latency/p99}}{{{p99}}}')
+            print(
+                rf'\drefset[unit=\unit{{\ms}}]{{{name}/latency/p99.9}}{{{p99_9}}}'
+            )
+            print(
+                rf'\drefset[unit=\unit{{\ms}}]{{{name}/latency/p99.99}}{{{p99_99}}}'
+            )
 
 
 def main():
@@ -232,12 +254,13 @@ def main():
                         nargs='+',
                         default=['line'])
     parser.add_argument("--summarize", action='store_true')
+    parser.add_argument("--dataref", action='store_true')
     parser.add_argument("--out", type=str, nargs='*', help='Output files')
 
     args = parser.parse_args()
 
     if args.summarize:
-        summarize(args.data_files, args.latency)
+        summarize(args.data_files, args.latency, dataref=args.dataref)
         sys.exit(0)
 
     plots = {