dsd versus no dsd

[riyaaora275]

If i have a code -

data test;
infile datalines ;
input a b c;
datalines;
1 2 3 
 4 5
6 7
;

I get two observations -

1 2 3

4 5 6

But if i use the code 

data test;
infile datalines dsd dlm=' ';
input a b c ;
datalines;
1 2 3
 4 5
6 7 
;

Here it identifies all the spaces as missing values.

result is 

1 2 3

. 4 5

6 7.

DSD identifies two delimiters as space where as in the code we have only one space before 4 and it is still identified as a delimiter and a missing value comes before 4.  Why?

[Kurt_Bremser · Accepted solution]

With dsd, each single delimiter character (blank since you did not specify another) is counted, so the second row in the datalines is considered as

<empty><blank>4<blank>5

and therefore supplies three values.

The third row is

6<blank>7<blank><emtpy>

and supplies another three values.

Without dsd, the leading blank in the second row is simply discarded, and the data step skips to a new line (reading 4/5/6 instead of ./4/5). Since it automatically skips to a new line at the end of the data step iteration (and no more lines are present), value 7 is never read and only 2 observations are output.

[riyaaora275]

In second line  you said the values are in this format

<empty><blank>4<blank>5

but isnt it in the below format because we have only one space before 4 and one space before 5. 

<blank>4<blank>5

where did  <empty> come from?

[Kurt_Bremser]

---

@riyaaora275 wrote:

In second line  you said the values are in this format

<empty><blank>4<blank>5

but isnt it in the below format because we have only one space before 4 and one space before 5. 

<blank>4<blank>5

where did  <empty> come from?

---

Because of the dsd option, the data step assumes that there has to be a value before the first delimiter (blank). If there's nothing (consider it a string of length 0), then you have a missing value. Without dsd, leading delimiters are skipped until there's something.