I might suggest using the DIVIDE function which will handle division by 0 and will assign special missing values for the division and not generate an error or warning message though in the Round function result you  will get missing values as a result.     ROUND(C * divide(100,/ A),0.01)   Or use the special missing .I  (yes there is period that is part of the value) instead to TEXT "inf" ... View more

You'd have to provide an example of your data. My guess is that your dates are already either date or datetime values and you simply need to add the desired format.   I suggest that you create a new thread and ask your question there.   Art, CEO, AnalystFinder.com   ... View more

While I agree entirely  with @RW9's observation that your  strategy is likely to make your project more inefficient, there is still a diagnosis to be made on the symptom you report.  Without seeing the macro code, I suspect that the SIZE parameter in macroA has to be less than 1000 for the last iteration of the loop whenever dataset TABLE1 does not have NOBS in the exact 1,000's.  If so, you can do this:   data _null_; set TABLE1 nobs=nobs ; file code ; nloops=int(nobs/1000); do part=0 to nloops; if part=nloops then size=mod(nobs,1000); else size=1000; put '%macroA(inputtable=TABLE1, outputtable = TABLE2, size=' size ',part=' part')' ; put '%union(inputtable = TABLE2, part = ' part')' ; end; stop; run; %include code / source2 ; You can test this on a  dataset with, say, 1011 observations. No need to wait for 442 iterations. ... View more

While I agree entirely  with @RW9's observation that your  strategy is likely to make your project more inefficient, there is still a diagnosis to be made on the symptom you report.  Without seeing the macro code, I suspect that the SIZE parameter in macroA has to be less than 1000 for the last iteration of the loop whenever dataset TABLE1 does not have NOBS in the exact 1,000's.  If so, you can do this:   data _null_; set TABLE1 nobs=nobs ; file code ; nloops=int(nobs/1000); do part=0 to nloops; if part=nloops then size=mod(nobs,1000); else size=1000; put '%macroA(inputtable=TABLE1, outputtable = TABLE2, size=' size ',part=' part')' ; put '%union(inputtable = TABLE2, part = ' part')' ; end; stop; run; %include code / source2 ; You can test this on a  dataset with, say, 1011 observations. No need to wait for 442 iterations. ... View more

You would have to tell use explicitly what values you change from and too. But unless you take extra steps that almost no one does if your original value was "2208,03" and you change the value to anything that might be interpreted as a number such as 220,803 then Excel changed the CELL to numeric. And very likely on file>save as>CSV (you did do this correct?) the value came out as ,220803, no quotes or even worse ,220,803, no quotes and now a different field.   Either edit with text editor, OR (sometimes this works, sometimes not) set the entire column in Excel as text before making any edit. OR use program code in a data step: If value='2208,03' <possibly with other qualifier such as: and ID='abc'> then value='220,803';   Also if you have multiple files with the same structure you would be much better off writing a data step because Proc Import guesses variable length and type every single time and your data sets may not be compatible. Note that Proc Import with CSV will write a data step to the log that you can copy, paste into the editor, modify and save. Then you change the infile and output data set to get the desired result.   You also should use guessingrows=  with a moderately large value or the key word MAX to have better compatibility on variable length when importing CSV or other delimited files. ... View more

@FP12 wrote: Thanks for the answer,   Nothing else than what I shared in my previous post in the log.     I modified what you suggesteed in  DBMS = xlsx   The import is working but then some column (not all) names are modified into "VAR1", "VAR2", ...   So my program is no longer working...   Thanks for the link, I will look at it later.  There are a number of possible causes for the "var1", "var2" variable name issues. One is what I call "phantom data". Excel will believe that there is a populated column if the file has ever had data in a cell in a column even if all the values are deleted. Proc import uses an engine that uses information provided by Excel and so you may get a variable for a column with no actual data. If you run proc freq on Var1 and all values are missing then this is a very likely occurrence. These are usually the right-most columns in the data. Note that you have also have entire rows of data with all values missing for the same reason.   Another cause is that the cell in the first row is blank. Sometimes you  may not notice this as Excel will flow text over from a cell to the left. Similar is having column headings that occupy two cells vertically for some rows of data. Though this often means that your data types are screwy as the second row of "header" becomes the first row of "data".   Yet another possibility is header text that is longer than 32 characters (the maximum that SAS will use for variable names) and exactly matches a column to the left. Proc Import will attempt to create unique variable names from column headers with identical text by appending 1, 2, 3 as needed for shorter text but when there are longer headings the algorithm doesn't really have a good handle and will default to "varX" names.   Please note that Proc Import is a guessing procedure and results may be quite odd with spreadsheets (Excel) likely to yield the oddest results. If you rely on Proc Import for multiple Excel files of supposedly the same format probability approaches unity that you will have inconsistencies in the resulting SAS data sets for either variable type (numeric/character/date valued) and/or length of variables and possibly variable names. With @RW9's suggestion you can have more control. ... View more

Use 8. informat in your INPUT function instead. ... View more

@FP12 unfortunately I don't think that's possible using the libname method.  ... View more

Another way to verify that the char var has been converted successfully, uses the double ? modifier in the input function. Thanks to @Kurt_Bremser for provding example data 😉   data test; input b :$32.; a = input(b, ?? best32.); if missing(a) and not missing(b) then put 'WARNING: Failed to convert ' b 'Check obs' _n_= ; format a best32.; cards; 123456.78 12345678912.22 1.12345678905 bob 3.14159265358979323846 ; run; proc print data=test noobs; run; ... View more

You are obviously skipping something very important as there are no values for B shown. The empty data set created by the shown SQL does have a length of 6. ... View more