For features I used on both, the average quality of the SQL Server implementation is between middling and pathetic (mostly the latter). I cannot recall a single instance where I thought "that's nice" or anything positive about how SQL Server does something. At best, it has been "this is not awful".
Things that sucked especially hard:
- Datetime API: In postgresql it is a joy to use and most features one would like to have to analyze data are directly there and work as they should. The analysis for my master's thesis was completely done in postgresql [1] (including outputting CSV files that could then be used by the LaTeX document) - without ever feeling constrained by the language or its features. Meanwhile, the way to store a Timespan from C# via EF is to convert it to a string and back or convert it to ticks and store it as long.
- The documentation: The pgsql documentation is the best of any project I have ever seen, the one for SQL Server is horrible and doesn't seem to have any redeeming qualities. - Management: SMSS is a usability nightmare and still the recommended choice for doing anything. I assume there are better ways but I haven't found them and if I have, they usually didn't work.
- The query language: In psql you can do things like
select a < b as ab, b < c as bc , count(*) from t group by 1,2; \crosstabview
select case when a < b then 1 else 0 end as ab, case when b < c then 1 else 0 end as bc, count(*) as count group by case when a < b then 1 else 0 end, case when b < c then 1 else 0 end
Maybe my use cases are a bit weird and I haven't encountered the better features of SQL Server yet. Would appreciate if someone could point out what I am doing wrong.
[1]: https://github.com/oerpli/MONitERO/blob/master/sql/queries/r...