Merge pull request #329 from openml/fix/task_dashboard

Fix Task Dashboard Top Score and Entry columns

server/src/dashboard/helpers.py

@@ -163,35 +163,6 @@ def get_data_metadata(data_id):
  return df, meta_features, numerical_features, nominal_features
 
 
-def get_highest_rank(df, leaderboard):
- df.sort_values(by=["upload_time"], inplace=True)
- scores = []
- # highest_rank = {}
- highest_score = {}
-
- setup_ids = []
-
- for index, row in df.iterrows():
- users = list(highest_score.keys())
- new_user = row["uploader_name"] not in users
- if row["setup_id"] not in setup_ids or new_user:
- setup_ids.append(row["setup_id"])
- score = row["value"]
- if new_user or (score not in scores):
- scores.append(score)
- scores.sort(reverse=True)
- # rank = scores.index(score) + 1
- if new_user or (highest_score[row["uploader_name"]] < score):
- # highest_rank[row['uploader_name']] = rank
- highest_score[row["uploader_name"]] = score
- # if highest_rank[row['uploader_name']] > row['Rank']:
- # highest_rank[row['uploader_name']] = row['Rank']
- # leaderboard['highest_rank'] = list(highest_rank.values())
-
- leaderboard["Top Score"] = list(highest_score.values())
- return leaderboard
-
-
 def splitDataFrameList(df, target_column):
  """df = dataframe to split,
  target_column = the column containing the values to split

server/src/dashboard/task_callbacks.py

@@ -10,7 +10,6 @@
 
 from .caching import CACHE_DIR_DASHBOARD
 from .dash_config import DASH_CACHING
-from .helpers import get_highest_rank
 from ...setup import SERVER_BASE_URL
 
 font = [
@@ -127,6 +126,7 @@ def update_task_plots(pathname, metric, n_clicks):
  ),
  )
  ]
+
  layout = go.Layout(
  autosize=False,
  margin={"l": 400},
@@ -158,7 +158,6 @@ def update_task_plots(pathname, metric, n_clicks):
  tick_text.append(link)
 
  df["upload_time"] = pd.to_datetime(df["upload_time"])
- df["upload_time"] = df["upload_time"].dt.date
 
  data = [
  go.Scatter(
@@ -177,7 +176,7 @@ def update_task_plots(pathname, metric, n_clicks):
  )
  ]
  layout = go.Layout(
- title="Contributions over time,<br>every point is a run, "
+ title="Contributions over time, every point is a run."
  "click for details",
  autosize=True,
  margin={"l": 100},
@@ -194,29 +193,22 @@ def update_task_plots(pathname, metric, n_clicks):
  fig1 = go.Figure(data, layout)
 
  # Leaderboard table
+ max_score_by_uploader = df[["uploader_name", "value"]].groupby(["uploader_name"]).max()
+ max_score_by_uploader = max_score_by_uploader.to_dict()["value"]
+ submissions_by_uploader = df["uploader_name"].value_counts().to_dict()
+ rank_by_uploader = {
+ uploader: rank
+ for rank, (uploader, score) in enumerate(sorted(max_score_by_uploader.items(), key=lambda t: -t[1]), start=1)
+ }
+
+ leaderboard = df.copy()[["uploader_name"]].drop_duplicates()
+ leaderboard["Entries"] = leaderboard["uploader_name"].apply(submissions_by_uploader.get)
+ leaderboard["Top Score"] = leaderboard["uploader_name"].apply(max_score_by_uploader.get)
+ leaderboard["Rank"] = leaderboard["uploader_name"].apply(rank_by_uploader.get)
+ leaderboard = leaderboard.rename(columns={"uploader_name": "Uploader"})
+ # Render order of columns matches dataframe column order.
+ leaderboard = leaderboard[["Rank", "Uploader", "Top Score", "Entries"]]
 
- top_uploader = df.sort_values("value", ascending=False).groupby(
- ["uploader_name"], sort=False
- )
- name = top_uploader["uploader_name"].unique()
- rank = list(range(1, len(name) + 1))
- entries = top_uploader["uploader_name"].value_counts().values
- leaderboard = pd.DataFrame(
- {"Rank": rank, "Name": name, "Entries": entries}
- ).reset_index()
- leaderboard.drop("Name", axis=1, inplace=True)
- ranks = []
- df = top_uploader.head(df.shape[1])
- for uploader in df["uploader_name"]:
- ranks.append(
- leaderboard[leaderboard["uploader_name"] == uploader].Rank.values[0]
- )
- df["Rank"] = ranks
-
- # Sort by time
- df.sort_values(by=["upload_time"], inplace=True)
- # Get highest score
- leaderboard = get_highest_rank(df, leaderboard)
 
  # Create table
  table = html.Div(