Language Composition Bestiary

Languages are generally composed using strings in Unix. This has well-known downsides like escaping problems leading to code injection attacks.

There are two ways that languages are composed using strings.

• One language can accept a string in another language (eval)
• One language can generate another: The most basic form of Metaprogramming

Eval (Runtime)

Shell:

• sh in sh: eval 'echo hi'
• sh in awk: system('ls | wc -l')
• sh in make
  • every single line in the actions in sh in make !!! .ONESHELL changes the semantics a bit.
  • $(shell find /) -- capturing it in a variable rather than an action.
• sh in Python/C, etc: os.system()

Make:

• make in make: $(eval $(mycode))

Other:

• X in sh. Shell is special, because more binaries have something like -e
  • python -c
  • perl -e
  • awk '{print $1}'
  • sed -e
• python in python: eval('1+2'), exec 'print "hi"'

Code Generation (Build Time)

• shell generating shell

• sed generating C

• awk generating C: TODO: Python build system had a sed example I rewrote in Awk. Shell

• C preprocessor generating C

• cmake generating Makefiles

• autoconf

  • m4 generating things that generate shell, make, C headers (config.h)
  • Makefile.pre.in, config.h.in, etc.

Web

• JS in HTML:
  • <script></script>
  • <onload="">
• CSS in HTML:
  • <style></style>
  • <p style="">
• HTML in JS:
  • document.innerHTML = '<p>hi</p>'
  • special extension: JSX, which is PHP-like
• CSS in JS:
  • element.style = 'font-size: large;'
  • document.querySelectorAll('#id')