About LaurieF

LaurieF · ‎10-30-2016

SAS does it all very nicely for you, with just a couple of additions, assuming I've understood you correctly. For a start, it makes things much easier if you store your date value as "days since 1 Jan 1960", or 20755 in this case. The easiest way to do this is to use the %sysevalf function: %let rundate = %sysevalf("28oct2016"d); /* which requires you to reformat your date, of course */ Alternatively: %let rundate = %sysfunc(inputn(161028, yymmdd6.)); Now, the intnx function is your friend. This will return the beginning, middle or end of a period ('month' in this case, obviously). Couple that with some formats and you're away. %let startofmonth = %sysfunc(intnx(month, &rundate, 0, b), date9.); %let endofmonth = %sysfunc(intnx(month, &rundate, 0, e), date9.); %let drundate = %sysfunc(putn(&rundate, monyy5.)); %put &startofmonth &endofmonth &drundate; This displays: 01OCT2016 31OCT2016 OCT16 Note that this will return 1OCT2016 etc. To convert that in macro code is a little ugly, but will be reliable: %let endofmonth = %substr(&endofmonth, 1, 3)%lowcase(%substr(&endofmonth, 4, 2))%substr(&endofmonth, 6, 4); If you want to see 1 October 2016 instead (which doesn't require the substringing): %let startofmonth = %sysfunc(intnx(month, &rundate, 0, b), worddatx.); Your other option, to return the year and month: %put %sysfunc(putn(&rundate, yymmdd4.)); 1610

LaurieF · ‎10-03-2016

I'm no great Perl maven, but using prxmatch will give you what you're after. Here's how I did it: data _null_; infile cards; retain x "GTTCACTAGCAACCTCAAACAGACACCATGGTGCACCTG"; length pattern $ 9; input pattern; match = prxmatch('/' || strip(pattern) || '/', x); put match=; cards; .CCTCA A.CTCA AC.TCA ACC.CA ACCT.A ACCTC. ; run; In Perl, the wild card is '.', not '-'. When this code is run, the pattern is found at position 12 every time. If it weren't found, like with index, it would return 0.

LaurieF · ‎09-13-2016

SAS Studio. I can get the XML mapping to work with a simpler, smaller file. And if I create an XML file from a SAS dataset and map that, it works. I was wondering if it were to do with the complexity of all the sub-files within the Health app export.

LaurieF · ‎09-13-2016

Hmm - I can't get this to work. I know I have write access to the folder, because I can create datasets and flat files in it programmatically; but there's no other hint as to what the error message might be. 57 filename health '/folders/myfolders/apple_health_export/export.xml'; 58 59 filename map '/folders/myfolders/apple_health_export/HealthData.map'; 60 61 libname health xmlv2 xmlmap=map automap=replace; ERROR: The creation of the XML Mapper file failed. ERROR: Error in the LIBNAME statement. <scratches head> Any ideas?

LaurieF · ‎08-19-2016

The straight programming style (not SAS-specific) was much more efficient: 5.59 seconds. %let max = 10000000; data _null_; length i 8; max=&max; array divisors[&max] 4 _temporary_; divisors[1]=1; do i=2 to max; divisors[i]=2; end; do j=2 to int(dim(divisors) / 2); do i=j * 2 to max by j; divisors[i] + 1; end; end; count = 0; do i = 2 to dim(divisors) - 1; count + divisors[i] = divisors[i + 1]; end; format max count comma14.; put 'There are ' count +(-1) ' consecutive positive divisors between 1 and ' max +(-1) '.'; stop; run; But not as much fun.

LaurieF · ‎08-19-2016

As many will attest, my mathematical abilities aren't great. My instincts aren't bad, but when I have to go beyond the delights of ones and zeroes, I'm prone to make mistakes. About six months ago I started hacking my way through https://projecteuler.net. I got stuck at about problem 30, and pretty much gave up. I was coding up the solutions in Apple's Swift - actually the reason was to teach myself to be a better Swift programmer. A couple of days ago I thought I'd have another go, and happened across problem 179 (https://projecteuler.net/problem=179), which reads: Find the number of integers 1 < n < 107, for which n and n + 1 have the same number of positive divisors. For example, 14 has the positive divisors 1, 2, 7, 14 while 15 has 1, 3, 5, 15. I wrote a solution which was taking ages (I gave up after thirty minutes), then I thought about it, recoded it and it came down to nine seconds. While I was out walking the dog today, I thought about writing it in SAS. When I came home I fired up SAS Studio on my Mac (I love being able to do that!) and hacked out the following code, using a hash table and 'by - notsorted'. The hash population takes 2'31", but the analysis steams through in four seconds. I'm ridiculously proud of: count + not(last.divisors); data _null_; length i 8; length divisors 4; max=1e7; dcl hash divhash(hashexp: log2(max), ordered: 'a'); divhash.definekey('i'); divhash.definedata('i', 'divisors'); divhash.definedone(); i=1; divisors=1; divhash.add(); do i=2 to max; divisors=2; /* All numbers > 1 have two divisors */ rc=divhash.add(); end; do j=2 to int(max / 2); do i=j * 2 to max by j; rc=divhash.find(); divisors + 1; rc=divhash.replace(); end; end; divhash.output(dataset: 'divisors'); stop; run; data _null_; set divisors nobs=max end=eof; by divisors notsorted; retain count 0; format max count comma14.; count + not(last.divisors); if eof; put 'There are ' count +(-1) ' consecutive positive divisors between 1 and ' max +(-1) '.'; run; There are always so many ways to skin a cat. I could have done it in one step by setting up an array instead of the hash. But I wanted to take advantage of by-group processing. You know the old criticism of "some people have too much time on their hands"? Sometimes I like being "some people".

LaurieF · ‎08-14-2016

It comes down to understanding what nodupkey does (and, as I just found out, nouniquekey!). The former detects multiple instances of duplicate keys and only keeps one of them. You have to be careful using this if you want to control which one is kept - if you do care, I recommend not using it; instead do a sort and then control which one in a data step. The latter only keeps duplicates. Running a sort with your data keeps the two observations you want for 033 but gets rid of the rest. So thank you - I learnt something new with nouniquekey.

LaurieF · ‎08-14-2016

There are so many ways to skin this particular cat (sorry, Lily), but nodupkey isn't one of them. It could have been done in one pass with SQL, but here's a method with proc sort and two data steps: data nodupkey_data; infile cards dsd dlm='09'x; attrib id length=3 format=z3.; attrib date informat=mmddyy6. format=yymmdd10.; attrib code length=$ 2; input id date code; cards; 009 123195 03 009 101595 01 009 090995 07 033 101095 07 033 101095 03 033 090995 02 033 051894 01 033 010594 07 045 122195 06 045 103195 05 045 102195 05 ; run; proc sort data=nodupkey_data; by id date; run; /* For each id, get the most recent date */ data maxdate; set nodupkey_data; by id; if last.id; keep id date; run; /* Merge back in with the main dataset, but only retain those observations where the date is the same as the most recent. */ data id_latest_date; merge nodupkey_data maxdate(in=in_maxdate); by id date; if in_maxdate; run;

LaurieF · ‎07-19-2016

Not quite - the original problem posited that the variables were separated by tabs, or '09'x. As long as that's the case, internal spaces are fine.

LaurieF · ‎07-19-2016

<nods> I had a chat with a SAS Institute chappy yesterday who pretty much said exactly what you have just put forward. Thank you very much. He apologised for the misdirection in the documentation. I left my X-terminal sessions on both the export and import machines running last night, and they both worked fine, because they both thought they had something to send stuff too, even though neither session ultimately wanted to. Heavy sigh. I'm going to chat to the Unix sysadmin today to see what he suggests. They might have a dummy DISPLAY value to use (the SAS guy said it was possible, but unlikely), but I will also ask about Xvfb. Laurie

LaurieF · ‎07-18-2016

When I use delimited infile code, I predefine all my variables with an attrib statement (or informat/format/length - but attrib is tidier). By not putting the length to be read in the input statement, it follows the delimiter control; the converse (as you have (understandably) done it) puts the cursor past the delimiter. Try this: data _null_; infile cards dsd dlm='09'x; attrib date informat=datetime20. format=datetime20.; input date num; put date= datetime20. num; stop; cards; 24-Jun-2012 03:39:00 79 ; run;

LaurieF · ‎07-18-2016

Yes, thank you, I saw that. But it doesn't answer my question; nor am I in a position to install more software on this machine. The documentation for Intelligence Platform batch tools implies that there is a setting for DISPLAY which will do what -disableX11 does in the third maintenance release. If I could work out why it works when I submit the ImportPackage command manually, but not when it's submitted by cron, then I might be closer to a solution.

LaurieF · ‎07-17-2016

I'm in the process of populating the DR box with metadata, and I've got the exporting, secure copy and importing all running tickety-boo. I started off using the wizard, then put the command it issued into a shell script and let cron take care of it. After I set the DISPLAY variable to my Unix session's localhost:12.0, it worked fine. But when I left it running over the weekend, it failed: "java.lang.InternalError: Can't connect to X11 winder server using 'localhost:12.0' as the value of the DISPLAY variable." I know there's a -disableX11 option in ImportPackage (which amusingly is described as being disabled by default: is it disabled or not?!), but that's not available in the version we're running. I've tried with other options I've searched for (variously localhost:0.0, :0.0, localhost:13.0), but really I'm fumbling in the dark. Does anyone know what I should set it to? why is the Unix session's value different to submitted-at-4am's requirement? why should they differ? Laurie

LaurieF · ‎05-17-2016

I agree that dictionary.tables (or sashelp.vtable) can be problematic, especially if the library contains views of external databases which it has to look through, because of columns like nobs. I've never, or rarely, had problems with dictionary.columns however. And f you're explicit about the libname and memname, the internal indexing is very efficient. Again in this case I knew the libname contained no views and it returned the response effectively instantly.

LaurieF · ‎05-17-2016

Quite so. However this is a fact table in a data warehouse, and thus all values (theoretically at least) are actual unformatted values. From the "more than one way of skinning a cat" department: another way of generating the SQL: proc sql noprint; select cat('max(', strip(name), ') as ', strip(name)) into :varstring separated by ', ' from dictionary.columns where libname = "%upcase(%scan(&_input, 1, .))" and memname = "%upcase(%scan(&_input, 2, .))"; create table &_output as select &varstring from &_input(&_input_options); quit;

Online Status	Offline
Date Last Visited	‎10-28-2024 09:00 PM