Merge pull request #85 from simvue-io/get_metrics

Adding metrics

Author: alahiff

CHANGELOG.md

@@ -2,6 +2,7 @@
 
 ## v0.13.0
 
+* Added methods to the `Client` class for retrieving metrics.
 * CPU architecture and processor obtained on Apple hardware.
 * Client now reports to server when files have been successfully uploaded.
 * `User-Agent` header now included in HTTP requests.

simvue/client.py

@@ -5,7 +5,7 @@
 
 from .serialization import Deserializer
 from .utilities import get_auth
-from .converters import to_dataframe
+from .converters import to_dataframe, metrics_to_dataframe
 
 CONCURRENT_DOWNLOADS = 10
 DOWNLOAD_CHUNK_SIZE = 8192
@@ -293,3 +293,67 @@ def get_folders(self, filters, tags=False, metadata=False):
             return response.json()
 
         raise Exception(response.text)
+
+    def get_metrics_summaries(self, run, name):
+        """
+        Get summary metrics for the specified run and metrics name
+        """
+        params = {'runs': run,
+                  'metrics': name,
+                  'summary': True}
+
+        response = requests.get(f"{self._url}/api/metrics", headers=self._headers, params=params)
+
+        if response.status_code == 200:
+            return response.json()
+
+        raise Exception(response.text)
+
+    def get_metrics(self, run, name, xaxis, format='list'):
+        """
+        Get time series metrics for the specified run and metrics name
+        """
+        params = {'runs': run,
+                  'metrics': name,
+                  'summary': False,
+                  'xaxis': xaxis}
+
+        if xaxis not in ('step', 'time', 'timestamp'):
+            raise Exception('Invalid xaxis specified, should be either "step", "time", or "timestamp"')
+
+        if format not in ('list', 'dataframe'):
+            raise Exception('Invalid format specified, should be either "list" or "dataframe"')
+
+        response = requests.get(f"{self._url}/api/metrics", headers=self._headers, params=params)
+
+        if response.status_code == 200:
+            if format == 'dataframe':
+                return metrics_to_dataframe(response.json(), xaxis, name=name)
+            return response.json()
+
+        raise Exception(response.text)
+
+    def get_metrics_multiple(self, runs, names, xaxis, sample_by, format='list'):
+        """
+        Get time series metrics from multiple runs and/or metrics
+        """
+        params = {'runs': ','.join(runs),
+                  'metrics': ','.join(names),
+                  'summary': False,
+                  'sample_by': sample_by,
+                  'xaxis': xaxis}
+
+        if xaxis not in ('step', 'time'):
+            raise Exception('Invalid xaxis specified, should be either "step" or "time"')
+
+        if format not in ('list', 'dataframe'):
+            raise Exception('Invalid format specified, should be either "list" or "dataframe"')
+
+        response = requests.get(f"{self._url}/api/metrics", headers=self._headers, params=params)
+
+        if response.status_code == 200:
+            if format == 'dataframe':
+                return metrics_to_dataframe(response.json(), xaxis)
+            return response.json()
+
+        raise Exception(response.text)

simvue/converters.py

@@ -44,3 +44,58 @@ def to_dataframe(data):
 
     df = pd.DataFrame(data=columns)
     return df
+
+def metrics_to_dataframe(data, xaxis, name=None):
+    """
+    Convert metrics to dataframe
+    """
+    import pandas as pd
+
+    if name:
+        columns = {xaxis: [], name: []}
+        for item in data:
+            columns[xaxis].append(item[0])
+            columns[name].append(item[1])
+
+        df = pd.DataFrame(data=columns)
+    else:
+        runs = []
+        metrics = []
+        for item in data:
+            if item[2] not in runs:
+                runs.append(item[2])
+            if item[3] not in metrics:
+                metrics.append(item[3])
+
+        headers = pd.MultiIndex.from_product([runs, metrics, [xaxis, 'value']], names=["run", "metric", "column"])
+
+        newdata = {}
+        for row in data:
+            if row[2] not in newdata:
+                newdata[row[2]] = {}
+            if row[3] not in newdata[row[2]]:
+                newdata[row[2]][row[3]] = []
+
+            newdata[row[2]][row[3]].append([row[0], row[1]])
+
+        max_rows = 0
+        for run in newdata:
+            for metric in newdata[run]:
+                if len(newdata[run][metric]) > max_rows:
+                    max_rows = len(newdata[run][metric])
+
+        results = []
+        for count in range (0, max_rows):
+            line = []
+            for run in newdata:
+                for metric in newdata[run]:
+                    if count < len(newdata[run][metric]):
+                        line.append(newdata[run][metric][count][0])
+                        line.append(newdata[run][metric][count][1])
+                    else:
+                        line.append(None)
+                        line.append(None)
+            results.append(line)
+
+        df = pd.DataFrame(data=results, columns=headers)
+    return df