Question

I looking to optimize an existing Makefile. It’s used to create multiple plots (using Octave) for every logfile in a given directory using an scriptfile for every plot which takes a logfilename as an argument. In the Moment, I use one single rule for every kind of plot available, with a handwritten call to Octave, giving the specific scriptfile/logfile as an argument.

It would be nice, if every plot has “his” octave-script as a dependency (plus the logfile, of course), so only one plot is regenerated if his script is changed.

Since I don’t want to type that much, I wonder how I can simplifiy this by using only one general rule to build “a” plot?

To make it clearer:

• Logfile: “$(LOGNAME).log”
• Scriptfile: “plot$(PLOTNAME).m” creates “$(LOGNAME)_$(PLOTNAME).png”

The first thing I had in mind:

%1_%2.png: %1.log
    $(OCTAVE) --eval "plot$<2('$<1')"

But this seems not to be allowed. Could someone give me a hint?

Answer 1

It’s pretty crazy that make doesn’t support this directly, I need it all the time.

The technique I use at the moment (with GNU make) (building on Didier Trosset’s example):

define OCT_template

all: %_$(1).png

%_$(1).png: %.log
    $$(OCTAVE) --eval "plot$(1)('$$*')"

endef

PLOT_NAMES = plot1 plot2 plot3

$(foreach p, $(PLOT_NAMES), \
  $(eval $(call OCT_template,$(p))) \
)

This is in the GNU make documentation.