How to write loop in a Makefile?

Viewed 320328

I want to execute the following commands:

./a.out 1
./a.out 2
./a.out 3
./a.out 4
.
.
. and so on

How to write this thing as a loop in a Makefile?

15 Answers

Although the GNUmake table toolkit has a true while loop (whatever that means in GNUmake programming with its two or three phases of execution), if the thing which is needed is an iterative list, there is a simple solution with interval. For the fun of it, we convert the numbers to hex too:

include gmtt/gmtt.mk

# generate a list of 20 numbers, starting at 3 with an increment of 5
NUMBER_LIST := $(call interval,3,20,5)

# convert the numbers in hexadecimal (0x0 as first operand forces arithmetic result to hex) and strip '0x'
NUMBER_LIST_IN_HEX := $(foreach n,$(NUMBER_LIST),$(call lstrip,$(call add,0x0,$(n)),0x))

# finally create the filenames with a simple patsubst
FILE_LIST := $(patsubst %,./a%.out,$(NUMBER_LIST_IN_HEX))

$(info $(FILE_LIST))

Output:

./a3.out ./a8.out ./ad.out ./a12.out ./a17.out ./a1c.out ./a21.out ./a26.out ./a2b.out ./a30.out ./a35.out ./a3a.out ./a3f.out ./a44.out ./a49.out ./a4e.out ./a53.out ./a58.out ./a5d.out ./a62.out

A simple, shell/platform-independent, pure macro solution is ...

# GNU make (`gmake`) compatible; ref: <https://www.gnu.org/software/make/manual>
define EOL

$()
endef
%sequence = $(if $(word ${1},${2}),$(wordlist 1,${1},${2}),$(call %sequence,${1},${2} $(words _ ${2})))

.PHONY: target
target:
    $(foreach i,$(call %sequence,10),./a.out ${i}${EOL})

Dynamically assign variables in the loop

The promblem with the for number in 1 2 3 4 ; do \ ...-solution is, that no variables can be assigned within the loop. $(eval VAR=...) can only be used, if the content of the assignment is known at the beginning of the target execution. If the assignment depends on the loop-variable, VAR will be empty.

To circumvent this issue, one can use the target functionality to model a loop. The following example takes the n-th file from SRC / OBJ and processes them together. Using this construction you can even use $(eval ...) to process the loop variable, as demonstrated with VAR3.

makefile

SRC = f1.c f2.cpp f3.cpp
OBJ = f1.o f2.o f3.o

SRC2 = $(addsuffix _,$(SRC))
JOIN = $(join $(SRC2),$(OBJ))

PHONY: all
all : info loop

loop : $(JOIN)

$(JOIN) :
    @# LOOP - CONTENT
    @echo "TARGET: $@"
    $(eval VAR1=$(word 1,$(subst _, ,$@)))
    @echo "VAR1: "$(VAR1)
    $(eval VAR2=$(word 2,$(subst _, ,$@)))
    @echo "VAR2: "$(VAR2)
    $(eval VAR3=$(subst .o,.x,$(VAR2)))
    @echo "You can even substitute you loop variable VAR3: "$(VAR3)
    #g++ -o $(VAR2) $(VAR1)
    @echo

PHONY: info
info:
    @printf "\n"
    @echo "JOIN: "$(JOIN)
    @printf "\n"

output

$ make

JOIN: f1.c_f1.o f2.cpp_f2.o f3.cpp_f3.o

TARGET: f1.c_f1.o
VAR1: f1.c
VAR2: f1.o
You can even substitute you loop variable VAR3: f1.x
#g++ -o f1.o f1.c

TARGET: f2.cpp_f2.o
VAR1: f2.cpp
VAR2: f2.o
You can even substitute you loop variable VAR3: f2.x
#g++ -o f2.o f2.cpp

TARGET: f3.cpp_f3.o
VAR1: f3.cpp
VAR2: f3.o
You can even substitute you loop variable VAR3: f3.x
#g++ -o f3.o f3.cpp

This answer, just as that of @Vroomfondel aims to circumvent the loop problem in an elegant way.

My take is to let make generate the loop itself as an imported makefile like this:

include Loop.mk
Loop.mk:Loop.sh
     Loop.sh > $@

The shell script can the be as advanced as you like but a minimal working example could be

#!/bin/bash
LoopTargets=""
NoTargest=5
for Target in `seq $NoTargest` ; do
    File="target_${Target}.dat"
    echo $File:data_script.sh
    echo $'\t'./data_script.ss $Target
    LoopTargets="$LoopTargets $File"
done
echo;echo;echo LoopTargets:=$LoopTargets

which generates the file

target_1.dat:data_script.sh
    ./data_script.ss 1
target_2.dat:data_script.sh
    ./data_script.ss 2
target_3.dat:data_script.sh
    ./data_script.ss 3
target_4.dat:data_script.sh
    ./data_script.ss 4
target_5.dat:data_script.sh
    ./data_script.ss 5


LoopTargets:= target_1.dat target_2.dat target_3.dat target_4.dat target_5.dat

And advantage there is that make can itself keep track of which files have been generated and which ones need to be (re)generated. As such, this also enables make to use the -j flag for parallelization.

This worked for me:

NUM=4

a-out:
    for (( i=1; i<=${NUM}; i++ )) \
    do \
        ./a.out $$i ; \
    done

Version

Code

files := $(wildcard ./*.txt ./**/*.go */**/*.js )

showFileFunc = echo "$(abspath ${1})\${2}"
delFileFunc = del "$(abspath ${1})\${2}"
cmdSplit = &
targetDisplay:
    $(foreach curFile, ${files}, ${call showFileFunc,${dir $(curFile)},${notdir $(curFile)}} ${cmdSplit})
targetDelete:
    $(foreach curFile, ${files}, ${call delFileFunc,${dir $(curFile)},${notdir $(curFile)}} ${cmdSplit})

Test Directory

Makefile
1.txt
 myDir
   - foo.go
   - bar.go
   - subDir
     - qoo.js

Test CMD & output

make showFile -s

output:
"C:/...\1.txt"
"C:/.../myDir\bar.go"
"C:/.../myDir\foo.go"          // since `.//**.js`
"C:/.../myDir/subDir\qoo.js"   // `.//**.js` can't but `*/**/*.js` is OK

Reference

Related