Fortran

Fortran (FORmula TRANslation) came out of IBM in 1957, and it is the oldest language on this site. It was built for numerical work and is still used for it, in weather and physics simulations, and in LAPACK, the linear-algebra library that NumPy and R build on. Here it is compiled with gfortran using the 2018 standard.

Modern Fortran works on whole arrays at once: x - mean subtracts from every element, and sum, minloc, dot_product, and matmul are built-in intrinsics. What it lacks is everything outside the numbers. There is no sort, no CSV reader, and no string splitting, so the implementations write their own. LAPACK itself is a separate library and isn’t installed here.

Quirks

• Without implicit none, a variable whose name starts with i through n is silently an integer: index_var = 2.9 stores 2. Every program here starts with implicit none, which makes an undeclared name a compile error.
• Arrays count from 1, and 2-D arrays are stored column by column.
• 7 / 2 is integer division, giving 3; 7.0 / 2 gives 3.5.
• The f0.6 format drops the zero before the point, printing 0.5 as .500000, so a small say helper puts it back.

git clone https://github.com/qnixon/dswithx.git
cd dswithx
python3 harness/run_tests.py --language fortran