Added basic integration with Lakeview Dashboards

pyproject.toml

@@ -20,7 +20,7 @@ classifiers = [
   "Programming Language :: Python :: Implementation :: PyPy",
 ]
 dependencies = [
-  "databricks-labs-blueprint>=0.4.2",
+  "databricks-labs-blueprint[yaml]>=0.4.2",
   "databricks-sdk>=0.22.0",
   "sqlglot>=22.3.1"
 ]

src/databricks/labs/lsql/dashboards.py

@@ -0,0 +1,82 @@
+import json
+from pathlib import Path
+from typing import Protocol, ClassVar, runtime_checkable
+
+import sqlglot
+import yaml
+from databricks.sdk import WorkspaceClient
+from databricks.sdk.service.workspace import ExportFormat
+
+from databricks.labs.lsql.lakeview import Dashboard, Page, Query, NamedQuery, ControlFieldEncoding
+
+@runtime_checkable
+class _DataclassInstance(Protocol):
+    __dataclass_fields__: ClassVar[dict]
+
+class Dashboards:
+    def __init__(self, ws: WorkspaceClient):
+        self._ws = ws
+
+    def get_dashboard(self, dashboard_path: str):
+        with self._ws.workspace.download(dashboard_path, format=ExportFormat.SOURCE) as f:
+            raw = f.read().decode('utf-8')
+            as_dict = json.loads(raw)
+            return Dashboard.from_dict(as_dict)
+
+    def save_to_folder(self, dashboard_path: str, local_path: Path):
+        local_path.mkdir(parents=True, exist_ok=True)
+        dashboard = self.get_dashboard(dashboard_path)
+        better_names = {}
+        for dataset in dashboard.datasets:
+            name = dataset.display_name
+            better_names[dataset.name] = name
+            query_path = local_path / f"{name}.sql"
+            sql_query = dataset.query
+            self._format_sql_file(sql_query, query_path)
+        lvdash_yml = local_path / "lvdash.yml"
+        with lvdash_yml.open('w') as f:
+            first_page = dashboard.pages[0]
+            self._replace_names(first_page, better_names)
+            page = first_page.as_dict()
+            yaml.safe_dump(page, f)
+        assert True
+
+    def _format_sql_file(self, sql_query, query_path):
+        with query_path.open('w') as f:
+            try:
+                for statement in sqlglot.parse(sql_query):
+                    # see https://sqlglot.com/sqlglot/generator.html#Generator
+                    pretty = statement.sql(
+                        dialect='databricks',
+                        normalize=True,  # normalize identifiers to lowercase
+                        pretty=True,  # format the produced SQL string
+                        normalize_functions='upper',  # normalize function names to uppercase
+                        max_text_width=80,  # wrap text at 120 characters
+                    )
+                    f.write(f"{pretty};\n")
+            except sqlglot.ParseError:
+                f.write(sql_query)
+
+    def _replace_names(self, node: _DataclassInstance, better_names: dict[str, str]):
+        # walk evely dataclass instance recursively and replace names
+        if isinstance(node, _DataclassInstance):
+            for field in node.__dataclass_fields__.values():
+                value = getattr(node, field.name)
+                if isinstance(value, list):
+                    setattr(node, field.name, [self._replace_names(item, better_names) for item in value])
+                elif isinstance(value, _DataclassInstance):
+                    setattr(node, field.name, self._replace_names(value, better_names))
+        if isinstance(node, Query):
+            node.dataset_name = better_names.get(node.dataset_name, node.dataset_name)
+        elif isinstance(node, NamedQuery) and node.query:
+            # 'dashboards/01eeb077e38c17e6ba3511036985960c/datasets/01eeb081882017f6a116991d124d3068_...'
+            if node.name.startswith('dashboards/'):
+                parts = [node.query.dataset_name]
+                for field in node.query.fields:
+                    parts.append(field.name)
+                new_name = '_'.join(parts)
+                better_names[node.name] = new_name
+            node.name = better_names.get(node.name, node.name)
+        elif isinstance(node, ControlFieldEncoding):
+            node.query_name = better_names.get(node.query_name, node.query_name)
+        return node

src/databricks/labs/lsql/lakeview/.codegen.json

@@ -0,0 +1,6 @@
+{
+  "formatter": "../../../../../../.venv/bin/black . && ../../../../../../.venv/bin/ruff . --fix",
+  "batch": {
+    "model.py.tmpl": "model.py"
+  }
+}

src/databricks/labs/lsql/lakeview/__init__.py

@@ -0,0 +1 @@
+from .model import *  # noqa: F401, F403