[prev in list] [next in list] [prev in thread] [next in thread] 

List:       sas-l
Subject:    Re: SAS Forum: Sum the B array columns based on corresponding non-zero values in the A array and by 
From:       Roger DeAngelis <rogerjdeangelis () GMAIL ! COM>
Date:       2018-06-30 10:19:55
Message-ID: CAOUdXL-k8hTPS2i8wx8A6vqab4hUwk=cw8V9csPTk=wrmtF_Yw () mail ! gmail ! com
[Download RAW message or body]

Good points

I suspect but do not know for sure.

  If the data is large, Teradata might sample the data and estimate the
skew in the grouping variable. It would use this skew information to assign
cores to groups. Might even create an index on the fly. I suspect it also
depends on the complexity of the logic. I think SAS does some of this.

On Fri, Jun 29, 2018 at 8:52 PM, Paul Dorfman <sashole@bellsouth.net> wrote:

> Pleasure, Roger.
>
> I must admit that I'm partial to the functionality of olde goode
> implicitly subscripted arrays (and their rather unique ability to
> incorporate other arrays as their elements) and like to mix the old and the
> new. That SAS doesn't "recommend" using them matters to me not a whit.
>
> SQL indeed doesn't require grouped/sorted data but more often than not it
> will sort behind-the-scenes. It might choose to hash when appropriately
> prompted but I don't believe it ever happens with data aggregation. Whether
> it is faster or not, it depends. If the optimizer should decide to sort,
> SQL will multi-thread the sort, yet MEANS will multi-thread as well, and so
> will the hash object in the DS2 environment.
>
> Btw, to me, SQL is "base" ;).
>
> Best regards
> Paul Dorfman
>
>

[Attachment #3 (text/html)]

<div dir="ltr"><div class="gmail_default" \
style="font-family:monospace,monospace">Good points</div><div class="gmail_default" \
style="font-family:monospace,monospace"><br></div><div class="gmail_default" \
style="font-family:monospace,monospace">I suspect but do not know for sure.</div><div \
class="gmail_default" style="font-family:monospace,monospace"><br></div><div \
class="gmail_default" style="font-family:monospace,monospace">   If the data is \
large, Teradata might sample the data and estimate the skew in the grouping variable. \
It would use this skew information to assign cores to groups. Might even create an \
index on the fly. I suspect it also depends on the complexity of the logic. I think \
SAS does some of this.</div></div><div class="gmail_extra"><br><div \
class="gmail_quote">On Fri, Jun 29, 2018 at 8:52 PM, Paul Dorfman <span \
dir="ltr">&lt;<a href="mailto:sashole@bellsouth.net" \
target="_blank">sashole@bellsouth.net</a>&gt;</span> wrote:<br><blockquote \
class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc \
solid;padding-left:1ex">Pleasure, Roger. <br> <br>
I must admit that I&#39;m partial to the functionality of olde goode implicitly \
subscripted arrays (and their rather unique ability to incorporate other arrays as \
their elements) and like to mix the old and the new. That SAS doesn&#39;t \
&quot;recommend&quot; using them matters to me not a whit.<br> <br>
SQL indeed doesn&#39;t require grouped/sorted data but more often than not it will \
sort behind-the-scenes. It might choose to hash when appropriately prompted but I \
don&#39;t believe it ever happens with data aggregation. Whether it is faster or not, \
it depends. If the optimizer should decide to sort, SQL will multi-thread the sort, \
yet MEANS will multi-thread as well, and so will the hash object in the DS2 \
environment. <br> <br>
Btw, to me, SQL is &quot;base&quot; ;).<br>
<br>
Best regards<br>
<span class="HOEnZb"><font color="#888888">Paul Dorfman<br>
<br>
</font></span></blockquote></div><br></div>



[prev in list] [next in list] [prev in thread] [next in thread]