An initial prototype of that in my local group-count branch quickly started running into problems: ```diff diff --git a/datasette/views/table.py b/datasette/views/table.py index be9e9c3..d30efe1 100644 --- a/datasette/views/table.py +++ b/datasette/views/table.py @@ -105,8 +105,12 @@ class RowTableShared(DataView): type_ = "integer" notnull = 0 else: - type_ = column_details[r[0]].type - notnull = column_details[r[0]].notnull + try: + type_ = column_details[r[0]].type + notnull = column_details[r[0]].notnull + except KeyError: # Probably count(*) + type_ = "integer" + notnull = False columns.append( { "name": r[0], @@ -613,6 +617,15 @@ class TableView(RowTableShared): offset=offset, )

• If ?_group_count we convert the SQL query here

• group_count = request.args.getlist("_group_count")
• if group_count:
• wrapped_sql = "select {cols}, count(*) from ({sql}) group by {cols}".format(
• cols=", ".join(group_count),
• sql=sql,
• )
• sql = wrapped_sql + if request.args.get("_timelimit"): extra_args["custom_time_limit"] = int(request.args.get("_timelimit")) Resulted in errors like this one: pk_path = path_from_row_pks(row, pks, not pks, False) File "/Users/simon/Dropbox/Development/datasette/datasette/utils/init.py", line 82, in path_from_row_pks bits = [ File "/Users/simon/Dropbox/Development/datasette/datasette/utils/init.py", line 83, in <listcomp> row[pk]["value"] if isinstance(row[pk], dict) else row[pk] for pk in pks IndexError: No item with that key ```

Had a new, different idea for how this could work: support a ?_group_count=colname parameter to the table view, which turns the page into a select colname, count(*) ... group by colname query - but keeps things like the filter interface, facet selection, search box and so on.

Implementing pagination for facets will be interesting. Would be easier if I had a nicer reusable internal pagination mechanism, which is also needed for #856 (pagination of canned queries).

Also relevant to this issue: #830 - redesigning the facet plugin hook in preparation for Datasette 1.0. And #972 supporting faceting against arbitrary queries.

For regular column faceting, here's the query that is used:

https://github.com/simonw/datasette/blob/13d1228d80c91d382a05b1a9549ed02c300ef851/datasette/facets.py#L196-L204

Since it uses order by count desc, value maybe those values could be used to implement cursor-based pagination.

That wouldn't be robust in the face of changing data, but I'm not sure it's possible to implement paginated faceting in a way that survives ongoing changes to the underlying data.

Maybe this ends up being code that defers to a simulated canned query, rendered using the existing query.html template.

URL design:

/database/table/-/facet/colname

And for other types of facet (to be supported later):

/database/table/-/facet/colname?_type=m2m