[ESQL] Support commands which dont need an index

[stratoula]

Describe the feature:

In ESQL we have 2 source commands:

• from , where you have to indicate your dataset
• row (i.e. ROW a = 1, b = "two", c = null ) where you can create a test dataset on your own

The first one creates succesfully an adhoc dataview and works great with Discover.
The second one no. The problem is here where we are creating a dataview with an empty name which Discover considers it invalid. As a result it renders an empty view.

We would like dataviews to also take under consideration this command. Another example is the show command.

For example: show functions. This also returns a datatable such as

{ "columns": [ { "name": "name", "type": "keyword" }, { "name": "synopsis", "type": "keyword" } ], "values": [ [ "abs", "abs(arg1)" ], [ "auto_bucket", "auto_bucket(arg1, arg2, arg3, arg4)" ], [ "avg", "avg(arg1)" ], [ "case", "case(arg1...)" ], [ "cidr_match", "cidr_match(arg1, arg2...)" ], [ "concat", "concat(arg1, arg2...)" ], [ "count", "count(arg1)" ], [ "count_distinct", "count_distinct(arg1, arg2)" ], [ "date_extract", "date_extract(arg1, arg2)" ], [ "date_format", "date_format(arg1, arg2)" ], [ "date_parse", "date_parse(arg1, arg2)" ], [ "date_trunc", "date_trunc(arg1, arg2)" ], [ "e", "e()" ], [ "is_finite", "is_finite(arg1)" ], [ "is_infinite", "is_infinite(arg1)" ], [ "is_nan", "is_nan(arg1)" ], [ "is_null", "is_null(arg1)" ], [ "length", "length(arg1)" ], [ "log10", "log10(arg1)" ], [ "max", "max(arg1)" ], [ "median", "median(arg1)" ], [ "median_absolute_deviation", "median_absolute_deviation(arg1)" ], [ "min", "min(arg1)" ], [ "mv_avg", "mv_avg(arg1)" ], [ "mv_concat", "mv_concat(arg1, arg2)" ], [ "mv_count", "mv_count(arg1)" ], [ "mv_dedupe", "mv_dedupe(arg1)" ], [ "mv_max", "mv_max(arg1)" ], [ "mv_median", "mv_median(arg1)" ], [ "mv_min", "mv_min(arg1)" ], [ "mv_sum", "mv_sum(arg1)" ], [ "now", "now()" ], [ "percentile", "percentile(arg1, arg2)" ], [ "pi", "pi()" ], [ "pow", "pow(arg1, arg2)" ], [ "round", "round(arg1, arg2)" ], [ "split", "split(arg1, arg2)" ], [ "starts_with", "starts_with(arg1, arg2)" ], [ "substring", "substring(arg1, arg2, arg3)" ], [ "sum", "sum(arg1)" ], [ "tau", "tau()" ], [ "to_bool", "to_bool(arg1)" ], [ "to_boolean", "to_boolean(arg1)" ], [ "to_datetime", "to_datetime(arg1)" ], [ "to_dbl", "to_dbl(arg1)" ], [ "to_double", "to_double(arg1)" ], [ "to_dt", "to_dt(arg1)" ], [ "to_int", "to_int(arg1)" ], [ "to_integer", "to_integer(arg1)" ], [ "to_ip", "to_ip(arg1)" ], [ "to_long", "to_long(arg1)" ], [ "to_str", "to_str(arg1)" ], [ "to_string", "to_string(arg1)" ], [ "to_ul", "to_ul(arg1)" ], [ "to_ulong", "to_ulong(arg1)" ], [ "to_unsigned_long", "to_unsigned_long(arg1)" ], [ "to_ver", "to_ver(arg1)" ], [ "to_version", "to_version(arg1)" ], [ "trim", "trim(arg1)" ] ] }

[elasticmachine]

Pinging @elastic/kibana-data-discovery (Team:DataDiscovery)

[stratoula]

This is causing the bug Lukas mention here #173913 (comment)

[stratoula]

We now don't return empty results but the user might encounter this bug if their previous index had the @timestamp column.

[kertal]

@mattkime @stratoula we should align here, because @mattkime is working on this: #179020

[stratoula]

I want this to be merged before the FF, I don't want to go GA with this bug folks. I haven't tested Matt's PR tbh. My PR is fixing this #180692 (unless I have missed something)

[kertal]

For sure! We will definitely not let this 🐛 to escape into the wilderness!