topic Re: Regex in SAS Programming

Regex

SASPhile — Thu, 11 May 2017 15:48:14 GMT

alive

is the subject alive

died

died at hospice

I would want to get only those records that contain only "alive" and "dead" as values. for instance row 1 and 3 are records to output. how do we get this using regex?

Re: Regex

RW9 — Thu, 11 May 2017 15:51:47 GMT

Why a regex, there is no pattern matching simply:

if upcase(variable) in ("ALIVE","DEAD") then ...

However note that died != dead, even though we can understand the two are the same, there is no logical way the computer can know.

Re: Regex

SASPhile — Thu, 11 May 2017 16:01:02 GMT

there are hunderds of words and later in time, the pattern matching may chnage and it is eaiser to have a regex

Re: Regex

RW9 — Thu, 11 May 2017 16:03:46 GMT

Sorry, still don't see it. Regex is for pattern matching, it wont really help you find words from a list. If there are lots then put them in a dataset thus:

data words;
  word="ALIVE"; output;
  word="DEAD"; output;
run;

proc sql;
  select  *
  from    HAVE
  where  YOUR_VARIABLE in (select * from WORDS);
quit;

Or you could do it in a format, or code generation, or by merging, but I don't see regex doing it.

Re: Regex

ballardw — Thu, 11 May 2017 16:18:26 GMT

@SASPhile wrote:

there are hunderds of words and later in time, the pattern matching may chnage and it is eaiser to have a regex

Your original post says:

"I would want to get only those records that contain only "alive" and "dead" as values"

That is exactly one word and confirmed by statement that rows 1 and 3 match (although whether you meant "died" instead of "dead" does come up).

With "only "alive" and "dead" " what does is matter if there are hundreds of words? You said "only".

It seems like you may not have clearly stated the problem or the desired result.

Re: Regex

kiranv_ — Thu, 11 May 2017 16:59:08 GMT

something like this in Regex should work. But what RW9 Suggests is the right way to do as it is much cleaner and very easily mangeable

data a;

input a $50.;

datalines;

alive

is the subject alive

died

died at hospice

Nothing happened

;

data c;

set a;

IF PRXMATCH("m/.*died.*|.*alive.*/oi", a) > 0 ;

run;

just add any word like .*wordyouwant.* after the pipe i.e. |. Here dot indicates any character and * means it can be there 0 or more times.

see this link for more explanation

http://support.sas.com/kb/38/719.html

Re: Regex

SASPhile — Thu, 11 May 2017 17:12:28 GMT

Tried this and worked out well!

^died\s*$|^alive\s*$/i

Re: Regex

kiranv_ — Thu, 11 May 2017 17:34:27 GMT

oh I think i did not read your question well. you want those words in the begining.

^died\s*$|^alive\s*$/i.

this will not catch if your variable value is like "alive is good". It can search for your find word with no space, single space or multiple spaces

if you try the below one just in case to be safe aand this takes care of your word at beginning+space after that+ any other other word(may or may not be there)

PRXMATCH("m/^died\s+.*|^alive\s+.*/i", a) > 0

Re: Regex

SASPhile — Thu, 11 May 2017 17:39:55 GMT

I want only those records contain just "died" and "alive"

Re: Regex

kiranv_ — Thu, 11 May 2017 17:59:56 GMT

for that your solution will work. you can also try this

PRXMATCH("m/^died$|^alive$/i", trim(a)) > 0

Re: Regex

Peter_C — Thu, 11 May 2017 18:22:46 GMT

I imagine a table of the words required with one row for each word string
Proc sql ;
Create table required_subset as
Select distinct a.*
From what.you_have a
Join imagined_words b
On a.description_variable contains trim( b.word_string )
;
quit ;
I used DISTINCT expecting there _might_ be multiple matches in some cases

Re: Regex

andreas_lds — Fri, 12 May 2017 13:00:12 GMT

@SASPhile wrote:

there are hunderds of words and later in time, the pattern matching may chnage and it is eaiser to have a regex

Do you know the required changes right now?

If not, why use regex? With "hunderds of words" the expression will be difficult to read. Using a format or the suggest proc sql is imho the smarter solution.

Re: Regex

Ksharp — Fri, 12 May 2017 13:13:25 GMT

IF PRXMATCH("/\b(died|alive)\b/oi", a) ;

Re: Regex

SASPhile — Wed, 31 May 2017 19:48:27 GMT

Hi ,

if there is a special character in the string and i want to match it as part of the string,how to include in expression?

if var1 has the following values:

var1

N-A

N/A

I would want to match NA and N/A only

Re: Regex

kiranv_ — Wed, 31 May 2017 20:11:21 GMT

something like this should work

'm/NA|N\/A/oi'

when you have N/A, you need to escape / by adding \ as / has different meaning

data abc;

input var1 $;

datalines;

N-A

N/A

;

data abcd(drop=pat);

set abc;

pat =prxparse('m/NA|N\/A/oi');

if prxmatch(pat, var1)>0 then newval ='yes';

else newval ='no';

run;

for your query all you need is

pat =prxparse('m/NA|N\/A/oi');

if prxmatch(pat, var1)>0 ;

Re: Regex

SASPhile — Fri, 09 Jun 2017 16:33:40 GMT

if the text conatins the key words discretion and physician , but the oder they appear varies. But I want to captuare only those records which have both discretion and phsycian in the text.

I can do a search by parsing the words in the order they might appear like as follows:

prxmatch("/(physician).*(discretion)|(discretion).*(physician)/oi",var_name).

i'm looking for alternate method where instead of passing it two times,is there a way I capture the record by parsing the key words just once?