Skip to content

thoughts.org

Latest commit

 

History

History
114 lines (82 loc) · 3.65 KB

thoughts.org

File metadata and controls

114 lines (82 loc) · 3.65 KB

Random Thoughts

Idea: TEF Processor Directive

A way to mark at the top of TEF files (and maybe other formats) that indicates a program that can process the file and translate it to some other format, like Turtle.

Sort of like a shebang line, but:

  • More than one can be specified
  • Programs are indicated in a host-agnostic way

What this header would say is something like

To translate this into a turtle document, pipe it into the program XYZ123 with arguments --foo=bar and --output-format=turlte.

It’s implied that the program’s standard input and output are used to read the TEF and write the converted output, respectively.

Which might look like…

tef:converters/to-turtle: XYZ123 --foo=bar --output-format=turtle

A standard command format would serve this purpose.

A language for commands

Syntax

See also: thoughts on ‘common syntax’ in TScript34’s README.

Supeficially similar to Bash or Tcl syntax.

  • Whitespace-separated tokens.
  • Backslash for escaping otherwise-special characters
  • Double quotes can be used to quote strings
  • Dollar sign is used for variable substitution and
  • Curly braces and parentheses can be used similar to how they are in Bash
  • Blank lines and lines

Some characters are reserved for extensions; when in doubt, quote tokens.

  • #, ' or `, * at the beginning of a token is reserved
  • Some single-character tokens are reserved:
    • ;
    • ( and )
    • { and }
    • [ and ]
  • Fancy unicode quotes also reserved. TOGVM-Spec and/or SchemaSchema may give examples of how each kind is to be interpreted.
    • Angle quotes seem to mean nestable, but no escape sequences
    • In general, quote types come in pairs, where single and double quotes have the same tokenization rules, but differ in semantics; single quoted text is treated as a symbol, whereas double-quoted text is treated as a literal string; in the case of this shell-like language that distinction would not make sense, since all text is literal and needs a $ to indicate anything else.

Commands are identified by URI. JCR36 and TScript34-P0019 define some already.

Some examples

Print “Hello, world!”:

http://ns.nuke24.net/JavaCommandRunner36/Action/Print "Hello, world!"

Same, but encoding the text to be printed as a URI:

http://ns.nuke24.net/JavaCommandRunner36/Action/Cat "data:,Hello,%20world!"

Tentatively

(which is why some things are ‘reserved’)

  • Unlike Bash Variable expansion happens after tokenization.
    • Use * to ‘splat’ a token foo *"bar baz" is the same as foo bar baz
  • Parentheses indicate sub-expressions
  • Square brackets make a list
  • Curly braces make a code block

‘lists’ and ‘code blocks’ in a language where all values are representable as strings (similar to Tcl) implies some canonical conversion.

So a hypothetical command like if { a ; b ; c } then { whatever } else { whatever-else } end would actually be converted to if some-encoding-of-a-then-b-then-c then some-encoding-of-whatever yaddah yaddah.

Make a project with test vectors

  • [ ] Basic commands (echo, exit)
  • [ ] Invoke Deno scripts
  • [ ] Execute URN-named Jar files.

For proper ‘scripting’

  • [ ] Define some ‘operator’ commands, like if, that take an entire program and run it

Proof-of-concept interpreter