aminghadersohi commented on PR #45486:
URL: https://github.com/apache/superset/pull/45486#issuecomment-6306593980
Acceptance run at e4a843beae51dfb87260a43cf25435c4368abd70, compared with
the merge base 7ff5709882fc9005da05a21654e3e2a167b29f70. Local Superset web +
MCP service, SQLite, example data (`birth_names`, yearly 1965–2008), filter `ds
>= '2000-01-01'`. Verdict: ship.
- MCP, 13 chart configs (Big Number trendline with/without grain, XY
line/bar/area/scatter with/without grain, mixed with/without grain, Waterfall,
pivot) × generate / unsaved preview / update_chart / update_chart_preview, as
Admin, Alpha and Gamma with dataset access: 168/168 cases return the filtered
data at head (9 rows, SUM(num) 12,347,764, the direct SQL total for ds ≥ 2000).
At the merge base 12/168 do (pivot only); the rest return all 44 years.
- Dashboard time binding via `extra_form_data`: a 2004–2008 window gives
6,296,981 at head and base; a 1990–2008 window keeps the user cutoff only at
head. Each query has a single time-range pair.
- Chart data API without MCP: queries without `granularity` are identical to
the merge base. With `granularity` (dashboard Time column filter), the merge
base dropped `>=`, `<` and `IS NOT NULL` comparisons on that column; head keeps
them.
- Browser: a native line chart and Big Number with time range 1990–2008 plus
`ds >= 2000-01-01`. Explore is the same at head and base. On a dashboard with
native Time range and Time column filters, the merge base plots 1990–2008 (the
chart's filter is dropped) and head plots 2000–2008, matching Explore.
- head:
https://github.com/aminghadersohi/superset/raw/pr-screenshots/apache-superset/45486/e4a843be/head_dashboard_1366x768.png
- base:
https://github.com/aminghadersohi/superset/raw/pr-screenshots/apache-superset/45486/e4a843be/base_dashboard_1366x768.png
- `query_dataset` with a dashboard-scope clause on the time column: the
merge base refused the call; head applies the clause (12,347,764).
No findings in the diff. One LOW finding outside the diff, already present
at the merge base: temporal filter values aren't validated as dates. `"2000"`
becomes `ds >= '1970-01-01 00:00:02'` (read as epoch ms), and `"not-a-date"`
reaches SQL unchanged. The MCP Big Number/XY paths now reach this because they
keep the filter. A follow-up could reject or normalize non-ISO values on
temporal columns in the MCP filter config.
Not covered: real LLM prompting, PostgreSQL, embedded guest, semantic views.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
