[Dashboard] Support more panels in dashboard

[abrarsheikh]

No description provided.

[gemini-code-assist]

Code Review

This pull request extends the dashboard functionality by adding support for more panel types and customization options, notably the 'table' panel. The changes introduce more flexibility in generating Grafana dashboards.

My review focuses on improving the maintainability and clarity of the new code:

• In grafana_dashboard_factory.py, I've suggested refactoring a large block of conditional logic to make it more readable and easier to maintain by using constant sets for panel types.
• In common.py, I've pointed out a confusing use of a duplicated ID in a new panel template and a misleadingly named parameter (heatmap_yaxis_label) that should be renamed for clarity.
• In test_metrics_head.py, I've suggested a small refactoring to remove duplicated code in a test case.

Overall, the changes are a good addition, and the suggested improvements will help with future development.

[gemini-code-assist]

The id for this TABLE_PANEL_TEMPLATE is 26. This ID is also used by GRAPH_PANEL_TEMPLATE, PIE_CHART_TEMPLATE, and BAR_CHART_PANEL_TEMPLATE. While the panel id is overridden during dashboard generation, using duplicate IDs in templates is confusing and could lead to subtle bugs if the generation logic changes. Please consider using a unique placeholder ID for each panel template to avoid potential conflicts.

[gemini-code-assist]

This large block of conditional logic has many repeated checks like panel.template in (...). This reduces readability and makes it harder to maintain.

To improve this, you could define constant sets for groups of panel templates that share configuration options at the top of the file.

For example:

PANELS_WITH_YAXES = {
    PanelTemplate.GRAPH,
    PanelTemplate.HEATMAP,
    PanelTemplate.STAT,
    PanelTemplate.GAUGE,
    PanelTemplate.PIE_CHART,
    PanelTemplate.BAR_CHART,
}

Then, you can simplify the conditions:

if panel.template in PANELS_WITH_YAXES:
    template["yaxes"][0]["format"] = panel.unit

Applying this pattern throughout this function will make the code more declarative and easier to manage.

[gemini-code-assist]

The logic to check targets is duplicated for row panels and other panels. You can refactor this to avoid repetition and improve readability.

                panels_to_check = []
                if panel["type"] == "row":
                    # Row panels contain nested panels, not targets directly
                    panels_to_check.extend(panel.get("panels", []))
                else:
                    panels_to_check.append(panel)

                for p in panels_to_check:
                    for target in p.get("targets", []):
                        assert serve_global_filters in target["expr"]

[cursor]

BAR_CHART panels ignore stack and linewidth settings

High Severity

The stack, linewidth, and nullPointMode settings are only applied when panel.template == PanelTemplate.GRAPH, but BAR_CHART panels also have these fields and need them applied. The BAR_CHART_PANEL_TEMPLATE has "stack": True and "linewidth": 1 as defaults, but several existing BAR_CHART panels in data_dashboard_panels.py explicitly set stack=False. Since the condition excludes BAR_CHART, these panels will incorrectly render as stacked when they should be grouped (unstacked). The condition should include PanelTemplate.BAR_CHART for stack, linewidth, and nullPointMode.

🔬 Verification Test

Test code:

# test_bar_chart_stack_bug.py
import copy
from ray.dashboard.modules.metrics.dashboards.common import (
    Panel, Target, PanelTemplate, BAR_CHART_PANEL_TEMPLATE
)
from ray.dashboard.modules.metrics.grafana_dashboard_factory import _generate_panel_template

# Create a BAR_CHART panel with stack=False (like the data dashboard panels)
panel = Panel(
    title="Test",
    description="Test",
    id=1,
    unit="short",
    targets=[Target(expr="test", legend="test")],
    stack=False,  # Explicitly setting stack to False
    template=PanelTemplate.BAR_CHART,
)

# Generate the template
result = _generate_panel_template(panel, [], 0, 0)

print(f"BAR_CHART_PANEL_TEMPLATE stack default: {BAR_CHART_PANEL_TEMPLATE['stack']}")
print(f"Panel.stack setting: {panel.stack}")
print(f"Generated template stack value: {result['stack']}")
print(f"BUG: stack=False was NOT applied: {result['stack'] == True}")

Command run:

cd /workspace && python test_bar_chart_stack_bug.py

Output:

BAR_CHART_PANEL_TEMPLATE stack default: True
Panel.stack setting: False
Generated template stack value: True
BUG: stack=False was NOT applied: True

Why this proves the bug: The panel was created with stack=False, but the generated template has stack: True (the default from BAR_CHART_PANEL_TEMPLATE). This proves the stack setting is being ignored for BAR_CHART panels, which will cause existing data dashboard histogram panels to render incorrectly as stacked instead of grouped.