Several count errors

[Pablolgn]

Initial Checks

• I have searched existing issues for duplicates
• I have read the documentation

Summary

I found some bugs some are already fixed in duckdb but fails in pandas.

DuckDB fixes 5 of the 8 bugs

Case	Pandas	DuckDB
[aggr n := count() group by Id_2]	NA, NA, NA, NA ❌	3, 3, 3, 1 ✅
analytic count(Me_int over (partition by Id_2)), group D	1 ❌	0 ✅
analytic order by without partition by	all 1 ❌	1…10 ✅
having + viral attribute	KeyError: ['Va_str'] 💥	3, 3, 3 ✅
analytic over String with nulls	TypeError: Invalid value '-1' 💥	3, 2, 2, 1 ✅

Present in both engines — 3 bugs

Case	Expected	Both
count(DS_1)	10	0
count(DS_1 group by Id_1)	3, 3, 3, 1	NA, NA, NA, NA
[aggr n := count(Me_int) group by Id_2], all-null group	0	NA

These three are the ones worth reporting: they are not backend-specific, they
are in the core. The first two are the same problem — count fails with a
dataset operand; with #component it works correctly on both engines.

Reproducible Example

DS_bug_1_ds_simple := count ( DS_1 );
DS_bug_2_ds_group  := count ( DS_1 group by Id_1 );
DS_bug_3_grupo_nulo := DS_1 [aggr n := count(Me_int) group by Id_2];
DS_bug_4_noarg     := DS_1 [aggr n := count() group by Id_2];
DS_bug_5_an_nulos  := DS_1 [calc n := count(Me_int over (partition by Id_2))];
DS_bug_6_an_sin_part := DS_1 [calc n := count(Me_num over (order by Id_2, Id_1))];
DS_bug_7_having    := DS_1 [aggr n := count(Me_num) group by Id_2 having count(Me_num) > 2];
DS_bug_8_an_string := DS_1 [calc n := count(Me_str over (partition by Id_2))];

Expected vs Actual

Expected: works fine

vtlengine version

1.9.2.rc4

Python version

No response

OS

No response

Additional Context

No response

[javihern98]

Check the behaviour for this issue given the count expected behaviour: #959 (comment)

[javihern98]

Implemented in #967