Sorting Through the Features of Proc Sort

[jaiganesh]

Hello,

Can anyone clarify on below concern,

"which one is good proc sql sorting or data sorting" ?

Regards,

Jai

[VDD]

SQL order by will have advantages as noted in this link

https://communities.sas.com/t5/SAS-Communities-Library/Custom-Sorting-with-Proc-SQL/ta-p/373460

normally straight ascending or descending sorts will give you what you want but just occasionally you will come up against a scenario where they won’t. In these cases, the flexibility afforded by the Proc SQL Order By clause can come to the rescue and turn what seems like a complex problem into something quite straightforward.

[PeterClemmensen]

Both good. Which one to use depends largely on the specific situation. You have to be more specific about your needs for us to give a more usable answer.

[ballardw]

Apples and oranges.

Proc sort has options to use different sort (collation) sequences, remove duplicate keys or observations to a separate output data set, prevent changing modified date on existing data, remove unique sort keys (only one record) and send those removed records to yet another data set.

If the ONLY operation being done on a data set Proc sort will usually be significantly faster.

If you actually doing what proc sql is designed for such as combining data sets such as joins and creating new variables, summarizind data or other manipulations then that is something that Proc sort doesn't do. So Sort would not be a valid option.

[novinosrin · Accepted solution]

"Apples and oranges."--Brilliant analogy. 

Beautifully explained with so much eloquence. First class!-->"If the ONLY operation being done on a data set Proc sort will usually be significantly faster.

If you actually doing what proc sql is designed for such as combining data sets such as joins and creating new variables, summarizing data or other manipulations then that is something that Proc sort doesn't do. So Sort would not be a valid option"

That's incredibly thorough. 

---

@ballardw wrote:

Apples and oranges.

Proc sort has options to use different sort (collation) sequences, remove duplicate keys or observations to a separate output data set, prevent changing modified date on existing data, remove unique sort keys (only one record) and send those removed records to yet another data set.

If the ONLY operation being done on a data set Proc sort will usually be significantly faster.

If you actually doing what proc sql is designed for such as combining data sets such as joins and creating new variables, summarizind data or other manipulations then that is something that Proc sort doesn't do. So Sort would not be a valid option.

---

[jaiganesh]

Thanks for fast and comprehensive reply.

[novinosrin]

That was @ballardw  's answer.