[SOLVED] Compute real number in CMake

Issue

I am using CMake to configure some scripts required to build my project using configure_file.

Some of these scripts do not have any math functionality, so I need to compute derived values in my CMake script using the math macro, e.g.:

math(EXPR EXAMPLE_BLOCK_SIZE "${EXAMPLE_SIZE} / ${EXAMPLE_BLOCK_COUNT}")

However, it seems that math does not support floating point arithmetic. If I set -DEXAMPLE_SIZE=2.5 and -DEXAMPLE_BLOCK_COUNT=2 CMake throws an error:

math cannot parse the expression: "2.5 * 2": syntax error, unexpected
exp_NUMBER, expecting $end (2)

Is there any way to compute a real number directly using CMake macros?

If not, what would be a portable way to achieve this?

Solution

First of all, floating-point math is not supported on CMake as far as I know. It is possible to compute variables, however, using shell commands. The following CMake function should do the trick:

function(floatexpr expr output)
    execute_process(COMMAND awk "BEGIN {print ${expr}}" OUTPUT_VARIABLE __output)
    set(${output} ${__output} PARENT_SCOPE)
endfunction()

Example usage:

set(A 2.0)
set(B 3.0)
floatexpr("${A} / ${B}" RESULT)
message(${RESULT})

NOTE: More advanced computations (such as trigonometric functions) can be computed using the command bc -l <<< 'EXPRESSION', but it's not part of the standard system packages and may not be installed.

EDIT: A completely portable (albeit much slower) solution would be to use the try_run functionality to compile a C file that runs the expression:

function(floatexpr expr output)
    SET(FLOAT_EXPR_SRC "
    #include <stdio.h>
    int main(int argc, char *argv[]){printf(\"%f\\n\", ${expr})\; return 0\;}
    ")
    FILE(WRITE ${CMAKE_CURRENT_BINARY_DIR}/CMakeFloatExpr.c ${FLOAT_EXPR_SRC})

    try_run(RESULT COMPRES ${CMAKE_CURRENT_BINARY_DIR}
            ${CMAKE_CURRENT_BINARY_DIR}/CMakeFloatExpr.c
            RUN_OUTPUT_VARIABLE __output)

    set(${output} ${__output} PARENT_SCOPE)
endfunction()

Answered By - Tal Ben-Nun