A quick tutorial on Miller

The input file has fields protected by double quotes, numbers with commas as decimal separator (Italian locale), and “-” in place of zeroes. A sample is available here and following table shows its content, verbatim

cat ./csv-cli-tools-input.csv

"name","surname","major","total","d1","d2"
"john","smith","arts","20","-","-16,8"
"carl","doe","engineering","30","-","12"
"adam","white","engineering","30","-","14,3"
"john","brown","arts","24","-","17,8"
"janet","smith","engineering","22","-","16,4"
"jane","doe","arts","20","14,8","26,8"
"anne","carol","engineering","10","12,2","26,3"

These are the tasks I had to perform on the file:

1. Pre-process the file to simplify analyses: convert separator in decimal numbers and replace “-” with zeroes
2. Compute descriptive statistics of “total”
3. Compute average total by major (pivot)
4. Transpose row and columns of the pivot table computed at the previous step
5. Select surname and total
6. Select all records with d1 < 13.0
7. Sort record by total

One of the issues with the file is the use of “,” both as field delimiter and separator of decimal places, which causes datamash to incorrectly parse the file.

We fix the format by:

• using sed to replace “-” with “0.00”
• using miller to convert the file to tsv, so that commas will appear only when used as decimal separators
• using sed to replace “,” with “.”, so that decimal numbers can be correctly interpreted as such

sed 's/"-"/"0.00"/g' csv-cli-tools-input.csv | mlr --icsv --otsv cat | sed 's/,/./g' | mlr --itsv --ocsv  cat > csv-cli-tools-input-cleaned.csv

Now we are ready to perform the analyses.

mlr --icsv --ocsv stats1 -a mean -f total -g major  csv-cli-tools-input-cleaned.csv | mlr --icsv --ocsv format-values -f "%.2f"

Note. Org Mode cleans data a bit by removing one zero from the second number.