WIP : dashboard

web_api/dashboard/app.py

@@ -0,0 +1,117 @@
+import os
+
+import pandas as pd
+import panel as pn
+from main_dashboard import MainDashboard
+from odmantic import SyncEngine
+from pymongo import MongoClient
+from ui import get_template
+
+from osm import schemas
+
+
+def flatten_dict(d):
+    """
+    Recursively flattens a nested dictionary without prepending parent keys.
+
+    :param d: Dictionary to flatten.
+    :return: Flattened dictionary.
+    """
+    items = []
+    for k, v in d.items():
+        if isinstance(v, dict):
+            # If the value is a dictionary, flatten it without the parent key
+            items.extend(flatten_dict(v).items())
+        else:
+            items.append((k, v))
+    return dict(items)
+
+
+def load_data():
+    if "LOCAL_DATA_PATH" in os.environ:
+        return pd.read_feather(os.environ["LOCAL_DATA_PATH"])
+    client = MongoClient(os.environ["MONGODB_URI"])
+    engine = SyncEngine(client=client, database="osm")
+    matches = (
+        engine.get_collection(schemas.Invocation)
+        .aggregate(
+            [
+                {
+                    "$match": {
+                        "osm_version": {"$eq": "0.0.1"},
+                        # "work.pmid": {"$regex":r"^2"},
+                        "metrics.year": {"$gt": 2000},
+                        # "metrics.is_data_pred": {"$eq": True},
+                    },
+                },
+                {
+                    "$project": {
+                        # "osm_version": True,
+                        # "user_comment": True,
+                        # "client.compute_context_id": True,
+                        "work.user_defined_id": True,
+                        "metrics.year": True,
+                        "metrics.is_code_pred": True,
+                        "metrics.is_data_pred": True,
+                        "metrics.affiliation_country": True,
+                        "metrics.score": True,
+                        "metrics.eigenfactor_score": True,
+                        "metrics.fund_pmc_anysource": True,
+                        "metrics.fund_pmc_institute": True,
+                        "metrics.fund_pmc_source": True,
+                        "metrics.journal": True,
+                    },
+                },
+            ]
+        )
+        .__iter__()
+    )
+    return pd.DataFrame(flatten_dict(match) for match in matches)
+
+
+def dashboard_page():
+    template = get_template()
+
+    dashboard = MainDashboard(pn.state.cache["data"])
+
+    template.main.append(dashboard.get_dashboard)
+    template.sidebar.append(dashboard.get_sidebar)
+
+    return template
+
+
+def on_load():
+    """
+    Add resource intensive things that you only want to run once.
+    """
+    pn.config.browser_info = True
+    pn.config.notifications = True
+    raw_data = load_data()
+    raw_data = raw_data[raw_data != 999999]
+
+    # Harcoded for now, will be added to the raw data later
+    raw_data["metrics"] = "RTransparent"
+
+    pn.state.cache["data"] = raw_data
+
+
+if __name__ == "__main__":
+    # Runs all the things necessary before the server actually starts.
+    pn.state.onload(on_load)
+    print("starting dashboard!")
+    pn.serve(
+        {"/": dashboard_page},
+        address="0.0.0.0",
+        port=8501,
+        start=True,
+        location=True,
+        show=False,
+        keep_alive=30 * 1000,  # 30s
+        autoreload=True,
+        admin=True,
+        profiler="pyinstrument",
+        allow_websocket_origin=[
+            "localhost:8501",
+            "osm.pythonaisolutions.com",
+        ],
+    )

web_api/dashboard/main_dashboard.py

@@ -0,0 +1,187 @@
+import holoviews as hv
+import pandas as pd
+import panel as pn
+import param
+
+pn.extension()
+
+
+pd.options.display.max_columns = None
+
+# filters = {
+#     "journal" : "category",
+#     "metrics" : "select",
+# }
+
+
+groups = {"year": "int"}
+
+
+dims_aggregations = {
+    "is_data_pred": ["percent", "count_true", "count"],
+    "is_code_pred": ["percent", "count_true"],
+    "score": ["mean"],
+    "eigenfactor_score": ["mean"],
+}
+
+
+aggregation_formulas = {
+    "percent": lambda x: x.mean() * 100,
+    "count_true": lambda x: (x == True).sum(),  # noqa
+    "count": "count",
+    "mean": "mean",
+}
+
+
+class MainDashboard(param.Parameterized):
+    """
+    Main dashboard for the application.
+    """
+
+    select_metrics = param.Selector(
+        default="RTransparent", objects=["RTransparent"], label=""
+    )
+
+    grouping_var = param.Selector(
+        default="year", objects=["year", "fund_pmc_institute"], label=""
+    )
+
+    filter_journal = param.Selector(
+        default="All journals (including empty)",
+        objects=[
+            "All journals (including empty)",
+            "All journals (excluding empty values)",
+            "Only selected journals",
+        ],
+        label="Journal",
+    )
+    filter_selected_journals = param.ListSelector(default=[], objects=[], label="")
+
+    def __init__(self, raw_data, **params):
+        super().__init__(**params)
+
+        self.raw_data = raw_data
+
+        self.param.filter_selected_journals.objects = self.raw_data.journal.unique()
+        # As default, takes the journals with the biggest number of occurences
+        self.filter_selected_journals = list(
+            self.raw_data.journal.value_counts().iloc[:10].index
+        )
+
+    def filtered_grouped_data(self):
+        filters = []
+
+        if self.filter_journal == "All journals (excluding empty values)":
+            filters.append(("journal.notnull()"))
+        elif self.filter_journal == "Only selected journals":
+            filters.append(f"journal in {self.filter_selected_journals}")
+
+        filtered_df = self.raw_data.query(*filters) if filters else self.raw_data
+
+        aggretations = {}
+        for field, aggs in dims_aggregations.items():
+            for agg in aggs:
+                aggretations[f"{agg}_{field}"] = (field, aggregation_formulas[agg])
+
+        result = (
+            filtered_df.groupby(self.grouping_var).agg(**aggretations).reset_index()
+        )
+
+        return result
+
+    @pn.depends("select_metrics", "filter_journal")
+    def get_sidebar(self):
+        items = [
+            pn.pane.Markdown("## Filters"),
+            pn.pane.Markdown("### Metrics extraction tool"),
+            pn.widgets.Select.from_param(self.param.select_metrics),
+            pn.layout.Divider(),
+            pn.pane.Markdown("### Grouping"),
+            pn.widgets.Select.from_param(self.param.grouping_var),
+            pn.layout.Divider(),
+            pn.pane.Markdown("### Filters"),
+            pn.widgets.Select.from_param(self.param.filter_journal),
+        ]
+
+        if self.filter_journal == "Only selected journals":
+            items.append(
+                pn.widgets.MultiChoice.from_param(
+                    self.param.filter_selected_journals, max_items=10
+                )
+            )
+
+        sidebar = pn.Column(*items)
+
+        return sidebar
+
+    @pn.depends(
+        "select_metrics", "filter_journal", "filter_selected_journals", "grouping_var"
+    )
+    def get_dashboard(self):
+        df = self.filtered_grouped_data()
+
+        # Create charts
+        fig_data_curve = hv.Curve(
+            df,
+            kdims=[self.grouping_var],
+            vdims=[
+                "percent_is_data_pred",
+            ],
+        ).opts(color="red")
+
+        fig_code_curve = hv.Curve(
+            df,
+            kdims=[self.grouping_var],
+            vdims=[
+                "percent_is_code_pred",
+            ],
+        ).opts(color="lightblue")
+
+        fig_data_points = hv.Points(
+            df,
+            kdims=[self.grouping_var, "percent_is_data_pred"],
+        ).opts(
+            tools=["hover"],
+            color="red",
+            size=5,
+            hover_tooltips=[
+                (self.grouping_var, f"@{self.grouping_var}"),
+                ("% is_data_prep", "@percent_is_data_pred"),
+                ("Total is_data_prep", "@count_true_is_data_pred"),
+                ("nbr_publications", "@count_is_data_pred"),
+            ],
+        )
+
+        fig_code_points = hv.Points(
+            df,
+            kdims=[self.grouping_var, "percent_is_code_pred"],
+        ).opts(
+            tools=["hover"],
+            color="lightblue",
+            size=5,
+            hover_tooltips=[
+                (self.grouping_var, f"@{self.grouping_var}"),
+                ("% is_code_prep", "@percent_is_code_pred"),
+                ("Total is_code_prep", "@count_true_is_code_pred"),
+                ("nbr_publications", "@count_is_data_pred"),
+            ],
+        )
+
+        plot = (
+            fig_code_curve * fig_data_curve * fig_data_points * fig_code_points
+        ).opts(
+            title="",
+            xlabel=self.grouping_var,
+            ylabel="Percentage",
+            width=800,
+            height=400,
+            legend_position="top_left",
+        )
+
+        # Layout the dashboard
+        dashboard = pn.Column(
+            "# Data and code transparency",
+            pn.Column(plot, sizing_mode="stretch_width"),
+        )
+
+        return dashboard

web_api/dashboard/ui.py

@@ -0,0 +1,71 @@
+import panel as pn
+
+
+def connection_monitor():
+    connection_monitor = pn.pane.HTML(
+        """
+
+    <script>
+
+        const originalSend = WebSocket.prototype.send;
+        window.sockets = [];
+        WebSocket.prototype.send = function(...args) {
+            if (window.sockets.indexOf(this) === -1)
+                window.sockets.push(this);
+            return originalSend.call(this, ...args);
+        };
+
+        console.log(window.sockets);
+
+        const polling = setInterval(function() {
+
+            if ( window.sockets.length > 0 ){
+
+                if ( window.sockets[0].readyState >= 2 ){
+
+                    let div = document.createElement('div');
+                    div.style.color = 'white';
+                    div.style.backgroundColor= 'crimson';
+                    div.style.padding = '10px 10px 10px 10px';
+                    div.style.textAlign= 'center';
+
+                    let text = document.createTextNode('Bokeh session has expired. Please reload.');
+                    div.appendChild(text);
+
+
+                    window.document.body.insertBefore(
+                        div,
+                        window.document.body.firstChild
+                    );
+
+                    clearInterval(polling);
+                }
+            }
+
+        }, 5000);
+
+    </script>
+    """
+    )
+
+    return connection_monitor
+
+
+def get_template():
+    """
+    Returns a Panel template with the given title,
+    with its menu and other header items.
+    """
+
+    template = pn.template.FastListTemplate(
+        site="NIH",
+        title="OpenSciMetrics",
+        favicon="https://www.nih.gov/favicon.ico",
+        sidebar=[],
+    )
+
+    template.header.append(
+        connection_monitor(),
+    )
+
+    return template