how to prevent rounding when converting character to numeric formats

[jdeere]

Attempting to convert character latitude/longitude values from character to numeric to utilize GEODIST function in PROC SQL.

Current code running on SAS (r) 9.4 (TS1M3)

data <dataset>;
set <tmp_dataset>;
LATITUDE2=input(LATITUDE, 20.15);
LONGITUDE2=input(LONGITUDE, 20.15);
run;

Output using PROC PRINT or as samples selected via PROC SQL displays as:

 AX_LATITUDE2  AX_LATITUDE  AX_LONGITUDE2  AX_LONGITUDE
 ------------------------------------------------------
     40.54115   40.541147        -105.077  -105.076775
      39.9924   39.992399        -104.978  -104.978418
     40.39116   40.391158        -104.712  -104.711740
     41.63468   41.634681         -93.698   -93.697993

[Tom]

You didn't post any example input so I am not sure what rounding you are talking about.

A couple of points.

Unless you know that your input character data does not include decimal places and you want SAS to insert one before the 15th to last character in the string then NOT include a the .15 on the INFORMAT.  If the data does have decimal places SAS will ignore the .15 but if the character value looks like an integer then adding the .15 will cause SAS to divide the value by 10**15.

8 byte floating point numbers cannot store 19 digits of precision. Look at the value of the function CONSTANT('exactint') for your platform to see what is the larger integer that SAS can store exactly.

Not all decimal values can be stored exactly using binary floating point numbers.

Attach a format to your numeric variable if you want it to print more digits. The default in most places is to use BEST12. or sometimes BEST8. to display numbers.

1172  data test;
1173    input @1 str $20. @1 num 20. ;
1174    best8=put(num,best8.);
1175    best12=put(num,best12.);
1176    best20=put(num,best20.);
1177    f20_15=put(num,f20.15);
1178    put (_all_) (=/);
1179  cards;


str=39.992399
num=39.992399
best8=39.9924
best12=39.992399
best20=39.992399
f20_15=39.992399000000000

str=40.391158
num=40.391158
best8=40.39116
best12=40.391158
best20=40.391158
f20_15=40.391158000000000

str=-93.697993
num=-93.697993
best8=-93.698
best12=-93.697993
best20=-93.697993
f20_15=-93.697993000000000

str=-104.711740
num=-104.71174
best8=-104.712
best12=-104.71174
best20=-104.71174
f20_15=-104.711740000000000

[TomKari · Accepted solution]

First, you're actually getting all of the digits, they're just not being displayed. Use a FORMAT statement to increase the number of digits displayed. See the program below.

Second, though, because of computing restrictions on 8 byte floating point representation, YOU CAN'T GET MORE THAN 15 SIGNIFICANT DIGITS accurately stored. You probably don't need that level of precision, but be aware.

Tom

data have;
length LATITUDE LONGITUDE $20;
LATITUDE = "40.5411471234567";
LONGITUDE = "-105.076775123456";
LATITUDE2=input(LATITUDE, 20.15);
LONGITUDE2=input(LONGITUDE, 20.15);
format LATITUDE2 LONGITUDE2 best20.;
run;

proc print;
run;

[jdeere]

that did the trick! Thanks!