C'est vraiment étrange. Trions plutôt dans l'ordre inverse :
select replace(replace('count(distinct <thiscol>) over (partition by <nextcol>) / count(*) over () as <thiscol>_<nextcol>,',
'<thiscol>', column_name
), '<nextcol>', lead(column_name) over (order by column_id desc)
)
from all_tab_columns atc
where table_name = 'mytable';
Notez le desc dans le tri.
