Importing an excel file from world bank with all data in a single column.

[nowak22]

Hey,

I was trying to import a CSV file with CPI data from World Bank, from here: https://data.worldbank.org/indicator/FP.CPI.TOTL.ZG

and importing a file named 'API_FP.CPI.TOTL.ZG_DS2_en_csv_v2_988671.csv'.

The problem I reckon is that there are few delimiters, including space, tab and comma.

Whatever I specify in a delimiter option I never get the expecting results. For instance: 

proc import datafile="/home/sasuser.v94/pliki/API_FP.CPI.TOTL.ZG_DS2_en_csv_v2_988671.csv"
        out=inflation_wb
        dbms=dlm
        replace;
        delimiter=' ,"';
        getnames=no;
run;

I will recieve an output with 74 columns and no data.

When I specify delimiter=' ,'; then I will get the ouput with all data packed in one column.

Anyone knows by any chance, how can I import this csv and get it properly displayed? 

[Kurt_Bremser]

Don't bother with proc import, write the data step yourself. You have country name, country code, two redundant columns you can drop anyway, and then numeric columns for 1960-2019. And the data starts in line 6, with line 5 holding the header, which might be hard for proc import to catch.

data want;
infile "......csv" dlm=',' dsd truncover firstobs=6;
input country_name :$40. country_code :$3. dummy :$1. dummy (_1960-_2019) (:20.);
run;

[Kurt_Bremser · Accepted solution]

And this would be my final code to create a proper dataset with minimal use of space and an intelligent structure for further processing (code is tested on University Edition):

data wide;
infile "/folders/myfolders/API_FP.CPI.TOTL.ZG_DS2_en_csv_v2_988671.csv" dlm=',' dsd truncover firstobs=6;
input country_name :$40. country_code :$3. dummy :$1. dummy (_1960-_2019) (:20.);
drop dummy;
run;

proc sort data=wide;
by country_code;
run;

proc transpose data=wide out=long (rename=(col1=income) where=(income ne .));
by country_code;
var _:;
run;

data cntlin;
set wide (
  keep=country_code country_name
  rename=(country_code=start country_name=label)
);
retain fmtname "country" type "C";
run;

proc format cntlin=cntlin;
run;

data want;
set long;
year = input(substr(_name_,2),4.);
format country_code $country.;
drop _name_;
run;

[nowak22]

Thank you kind Sir, I will have to work on your code then to fully understand it.

[Kurt_Bremser]

---

@nowak22 wrote:

Thank you kind Sir, I will have to work on your code then to fully understand it.

---

The first step reads the file into a dataset; to make it easier to code, I used a variable list (_1960-_2019), which creates all variables in the sequence automatically.

Next, I sorted this dataset to facilitate the use of BY in the transpose step.

The transpose converts the horizontal structure to a vertical structure, removes the country name, and removes all missing values by means of a WHERE= dataset option.

Then I extract the country code and country name and rename them, add the necessary variables for a "controlin" dataset for PROC FORMAT, and execute PROC FORMAT to create a format that maps from country code to country name.

In the final step, I ceate a numeric YEAR variable from the variable _NAME_, which was automatically created in the transpose step. In this step, I also use the newly created format.

If the resulting dataset is to be used later, you should add the format to a permanent format catalog that is in your FMTSEARCH system option.