makefile: from 3 input generate one output - makefile

Question

I have this version of makefile
[sbsuser#compute-00-01 415]$ make --version GNU Make 3.81
I have directory SOMATIC where I have 3 file . I want to produce a only one output. This is what I wrote.
`
OUTSOMATIC=SOMATIC
FINAL=FINAL
INPUT=$(wildcard $(OUTSOMATIC)/*.vcf)
OUTSORT2= $(patsubst $(OUTSOMATIC)/%.vcf,$(FINAL)/%somatic.ensemble.gz,$(INPUT))
$(info lista $(OUTSORT2))
$(info lista $(INPUT))
.PHONY: all
all: $(INPUT) $(OUTSOMATIC) $(OUTSORT2) $(FINAL)
$(FINAL)/%somatic.ensemble.gz: $(OUTSOMATIC)/%.vcf $(INPUT)
~/jdk1.8.0_121/bin/java -XX:+UseSerialGC -Xms1g -Xmx10g -jar /illumina/software/PROG2/bcbio-variation-recall-0.1.7 ensemble -n 1 $(FINAL)/somatic_ensemble.gz /illumina/software/database/database_2016/hg19_primary.fa $^
`
With this script make 3 time the same files. I don't understand how to create only one output from list of input to use in the same time.
What is the best way to do this?
If I change $(FINAL)/%somatic.ensemble.gz: in $(FINAL)/somatic.ensemble.gz I have this error:
make: *** No rule to make target FINAL/415_merge_mutect2.somaticsomatic.ensemble.gz', needed byall'. Stop`

Answer 1

You probably should review the GNU make manual introductory sections where they describe how make works.
Let's look at your makefile; first you define some variables. Let's assume that you have the files SOMATIC/foo.vcf, SOMATIC/bar.vcf, and SOMATIC/baz.vcf. Then the variables you created will have these values, after they are expanded:
OUTSOMATIC = SOMATIC
FINAL = FINAL
INPUT = SOMATIC/foo.vcf SOMATIC/bar.vcf SOMATIC/baz.vcf
Now your patsubst finds all words in INPUT that match the pattern SOMATIC/%.vcf and replace that with FINAL/%somatic.ensemble.gz, where the part that matches the % in the input is substituted into the output:
OUTSORT2 = FINAL/foosomatic.ensemble.gz FINAL/barsomatic.ensemble.gz FINAL/bazsomatic.ensemble.gz
Now, make sees that you've defined an all target. Since it's the first target in the makefile this is the target that will be run by default. After expansion, it will look like this:
all: SOMATIC/foo.vcf SOMATIC/bar.vcf SOMATIC/baz.vcf SOMATIC FINAL/foosomatic.ensemble.gz FINAL/barsomatic.ensemble.gz FINAL/bazsomatic.ensemble.gz FINAL
So, make will try to build every prerequisite of the all target to be sure it's up to date. First it tries to build the SOMATIC/*.vcf files. Those files already exist and make doesn't have any rules about how to rebuild them, so it assumes they're up to date.
Next it tries to build the SOMATIC file. This is a directory and it also has no rule to be built, so make assumes that's up to date as well.
Next make tries to build the target FINAL/foosomatic.ensemble.gz. Make does have a rule that can build it, you've created one:
$(FINAL)/%somatic.ensemble.gz: $(OUTSOMATIC)/%.vcf $(INPUT)
~/jdk1.8.0_121/bin/java ...
This matches the target you want to build, with a % value of foo, so then make substitutes the % in the prerequisite for foo and finds that SOMATIC/foo.vcf exists and doesn't need to be rebuilt, so it runs your recipe. However your recipe doesn't actually create the target FINAL/foosomatic.ensemble.gz; it creates the target FINAL/somatic_ensemble.gz. So this rule is broken because it tells make it will do one thing, but it does something else.
You should always ensure all your recipes build the file represented by the automatic variable $#; that will ensure that you and make agree on the meaning of your rule. If you want your recipe to build some other file, then your rule is written incorrectly.
Next make does the same thing with the next prerequisite of all: FINAL/barsomatic.ensemble.gz. Since that file doesn't exist, make tries to build it using the pattern rule, but again that creates the same output file.
And again for the third .gz file FINAL/bazsomatic.ensemble.gz. That's why things are run three times.
If you change the pattern rule to an explicit rule building FINAL/somatic.ensemble.gz, which is what you want, then make can't find any way to build the prerequisites of the all target so it gives this error.
Your problem is the creation of OUTSORT2. You want to create only one output file, but you've set OUTSORT2 to contain three different files, so make tries to create all three files. You want this:
OUTSOMATIC = SOMATIC
FINAL = FINAL
INPUT = $(wildcard $(OUTSOMATIC)/*.vcf)
OUTSORT2 = $(FINAL)/somatic.ensemble.gz
.PHONY: all
all: $(OUTSORT2)
$(OUTSORT2): $(INPUT)
~/jdk1.8.0_121/bin/java -XX:+UseSerialGC -Xms1g -Xmx10g -jar /illumina/software/PROG2/bcbio-variation-recall-0.1.7 ensemble -n 1 $# /illumina/software/database/database_2016/hg19_primary.fa $^