A naive evaluator for the functional programming language FL as described in FL Language Manual, Parts 1 and 2 (Backus et al., 1989) [https://theory.stanford.edu/~aiken/publications/trs/RJ7100.pdf].

"Naive" means that it is too slow for a practical interpreter, and, more importantly, there is no tail recursion optimization. Therefore the JVM stack size must be chosen large enough for the expression to be evaluated.

fleval [ OPTION ]... [ FILE ]

Evaluates the FL expression in FILE. When FILE is -- reads expression from standard input.

When FILE is omitted, the interactive mode is started.

• -p : In non-interactive mode, the result of the evaluation is printed on standard output.
• -L PATH : PATH is added to the search paths for library environments. A library file lib.fl is first looked up in the current working directory, then in the directories of the search paths in order specified using the -L options.
• -c : Only the syntax of FILE is checked, it is not evaluated.
• -to-fl : Standard input is converted to the FL XML data format and printed on standard output.
• -from-fl : FL XML data format is read from standard input and converted to text on standard output. This only works if the data consists of a single string element.
• -V or --version : Prints the current version of the evaluator.
• -h- or --help : Displays the help.

In the interactive mode, expressions are entered into a REPL at the prompt :. Expressions are evaluated on pressed ENTER, except when the brackets (, [, <, [| are not balanced. At the next prompt, each unbalanced bracket is displayed before the prompt. A line containing the single command :) close all unbalanced brackets. The command :u FILE adds the library in FILE to the standard environment. The functions exported by FILE can then be used, as if they were primitive functions.

The emacs supports highlighting of FL expressions and input of special characters. Automatic indenting is currently not implemented.

See also [emacs/README.txt].

The in primitive reads a character from standard input, if the device name is kbd (buffered). If the device is a file name, then the file with this name is opened once for reading. Subsequent calls to in with the same name read the next character from the already opened file.

The out primitive writes a string to standard output, if the device name is scr. If the device is a file name, then the file with this name is opened once for writing. Subsequent calls to out with the same name append to the already opened file.

The get and put primitives read and write FL values in an XML data format. Only values made of atoms and sequences are supported.

See also the RelaxNG schema for the FL XML data format in [doc/fleval.rnc].

• The character / starts an identifier, thus a/b must be written a / b, since /b is parses as an identifier,
• The characters + and - start a number, this a-2 must be written a - 2, and a + 2 must be written a + 2.
• Similar rules apply to <- and ->, <=| and |=>, [| and |]. For example, the sequence containing the function - must be written < - > instead of <->.

• Identifiers may start with a letter from a to z, A to Z, the characters /, $, %, #, _, ?, ↑, ↓, ^, ¬, as well as Greek letters from Α to ω (Unicode range 0x0391 to 0x03C9).

  Their names may continue with the above and the digits 0 to 9.

• For the character and string escape sequences, see [doc/escapes.txt].

• The name alpha is an alias for the primitive function α, the name Delta is an alias for the primitive function Δ.

• No tail recursion optimization is performed (yet).

• In the rule exdef pat ... , pat must be a simple expression, in order to avoid an ambiguous grammar. Complex expressions can still be used by putting them within parentheses.

• Used defined pattern combiners are not supported.

• Error handling while evaluating environments is patchy.

• Nested comments are not supported.

• The implementation of the most general form of definitions is questionable.