Executing a function per month in a date range

Question

Problem: For parameters of a minimum and maximum date (ex: script.sh 20190801, 20201005) execute a function on each month within that date range. (ex: 20190801-20190831, 20190901-20190931 . . .)

Backstory: I have a function that generates a CSV based on data from a date range, however for large date ranges this takes too long to run. That in mind, I'm attempting to have this run per-month simultaneously and stitch the CSVs back together afterwards. I would attempt something like this in Python or Javascript and use a datetime library, but certain company limitations are requiring me to exclusively use bash/shell.

Below is my current idea, but it fails in most cases

I get my expected results from passing, for example: 20190101 20201231 to the below code, but I'm unsure how to modify this to support something like 20190105 20200820

DateMin=$1
DateMax=$2

if [ $DateMin -ge $DateMax ]; then
    echo "ERROR: Minimum Date Cannot be above Maximum Date"
    exit
fi

let DateDelta=DateMax-DateMin

for YearDelta in `seq $DateDelta -10000 0`
do
    let TempDelta=$YearDelta-1130
    for MonthDelta in `seq $YearDelta -100 $TempDelta`
    do
        let TempMin=$DateMax-$MonthDelta
        let TempMax=$DateMax-$MonthDelta+30
        call_to_function $TempMin $TempMax
    done
done

I created a repl.it here so it's easier to see: https://repl.it/repls/CanineAdmirableSeahorse#main.sh If you run "bash main.sh 20190101 20201231" that's the expected output.

Answer 1

How about using date program, which is there, and can calculate a lot with dates. Here is an example:

DateMinStr="${1:0:4}-${1:4:2}-01"
DateMaxStr="${2:0:4}-${2:4:2}-${2:6:2}"
DateMin=$(date -d $DateMinStr +%s)
DateMax=$(date -d $DateMaxStr +%s)

if [ $DateMin -ge $DateMax ]; then
    echo "ERROR: Minimum Date Cannot be above Maximum Date"
    exit
fi

DateCurrentStr=$DateMinStr
DateCurrent=$DateMin
while [ $DateCurrent -lt $DateMax ]
do
    TempMin=$(date -d $DateCurrentStr +%Y%m%d)
    TempMinStr=$(date -d $DateCurrentStr +%Y-%m-%d)
    TempMax=$(date -d "$TempMinStr + 1 month - 1 day" +%Y%m%d)
    TempMaxStr=$(date -d "$TempMinStr + 1 month - 1 day" +%Y-%m-%d)
    echo $TempMin $TempMax
    DateCurrent=$(date -d "$TempMaxStr + 1 day" +%s)
    DateCurrentStr=$(date -d "$TempMaxStr + 1 day" +%Y-%m-%d)
done

I don't think this example does exactly what you want, and is far from optimized, but gives you an idea how to do it.

Answer 2

Tools like dateutils are usually better suited for this kind of work but it can be done with shell primitives, here with a different set of parameters than yours, for brevity.

#!/bin/sh
# For a date range print start and end date per calendar month.

# Gregorian calendar: number of days in month.
# $1 year (>= 1600)
# $2 month (1..12), one leading '0' allowed for 1..9
# exitval: day count (28..31)
daysinmonth() {
    case "$2" in
    (2|02)
        test 0 -eq $(($1 % 4 || !($1 % 100) && $1 % 400))
        return $((28 + ($?<1))) ;;
    (4|6|9|11|04|06|09)
        return 30 ;;
    (*) return 31 ;;
    esac
}

# Print year,month,day in ISO 8601 basic format
isobasic() {
    printf '%d%02d%02d' "$1" "$2" "$3"
}

ylo="$1" mlo="${2#0}" dlo="${3#0}"      # range start
yhi="$4" mhi="${5#0}" dhi="${6#0}"      # range end
isohi=$(isobasic "${yhi}" "${mhi}" "${dhi}")
until 
    isolo=$(isobasic "${ylo}" "${mlo}" "${dlo}")
    test $(( isolo - isohi )) -gt 0
    # test "${isolo}" \> "${isohi}"     # dash, bash
do
    # compute end of month
    daysinmonth "${ylo}" "${mlo}"
    isoeom=$(isobasic "${ylo}" "${mlo}" "${?}")
    # print start, end date for partial/whole calendar month
    test "${isoeom%??}" != "${isohi%??}" || isoeom="${isohi}"
    printf '%s%s %s\n' "${PREFIX:-}" "${isolo}" "${isoeom}"
    # increase *lo values to first of next month
    dlo=1
    : $(( mlo = (mlo == 12 ? 1 : mlo + 1) ))
    : $(( ylo = (mlo > 1 ? ylo : ylo + 1) ))
done
# fail if start later than end date
return $(( isoeom == isohi ? 0 : 1 ))

Given the command

PREFIX='cmd ' ./script 2016 11 09 2020 3 5

output is:

cmd 20161109 20161130
cmd 20161201 20161231
cmd 20170101 20170131
cmd 20170201 20170228
...
cmd 20191201 20191231
cmd 20200101 20200131
cmd 20200201 20200229
cmd 20200301 20200305