samtrosku specification

From Lojban
Jump to: navigation, search

samtrosku is a procedural computer language expressed in Lojban.

la samtrosku xi no pi'e pa pe fi'e la rab.spir

A samtrosku program must parse. Certainly some parts of the language could be simplified for computers - for example, names wouldn't necessarily have to be pronounceable - but samtrosku's unique advantage is that it could be spoken aloud. The commands and functions in samtrosku are brivla, and are defined by their place structure.

Statements and blocks

  • A statement is a bridi.
  • A statement can be a simple bridi such as li abu mleca li ci. This simply returns a truth value of "true" or "false", which can be used if the next statement is connected with a logical connective, or if the statement is part of a function.
  • A statement can also be a bridi involving ko. This is a command for the computer to do something, such as ko cusku lu coi munje li'u. It returns "true" if the command can be performed, and "false" otherwise. Because of the way truth values work, {jvs|ko}} is always equivalent to do.
  • If a statement is connected to another statement with a logical connective, the first statement is performed, and then the second statement is either performed or skipped in order to make the connective true. If the connective would be true (or false) either way, the second statement is skipped - call this the "laziness rule". The text scope containing both statements is assigned a truth value - if the connective cannot be made true, the truth value of the connected statements is false; otherwise it is true.
  • If the end of a text scope is reached (there are no more connectives, and the commands are not inside a block) and the last truth value is "false", the program stops with an error. So a command which cannot be executed will stop the program. Also, this means a simple statement, not connected to anything, can be used as an assertion - if it is true, nothing happens, and if it is false, the program stops.

Blocks

  • A block is a set of statements grouped with tu'e ... tu'u. It creates a separate text scope, which is true if all the statements within are true, and false otherwise (at which point no more statements in the block are executed). Unlike a logical connective scope, a block does not overwrite go'u and go'a (see below).
  • The cmavo di'e would refer to whatever is in the following text scope (but must be used as la'e di'e to actually run the code). So di'e would most commonly be used to introduce a block.
  • Blocks are useful for defining loops and brivla.

Recent truth values

If you want to do something with logical values that connectives cannot handle (such as an if-then-else statement), you can use the go'V series of cmavo.

  • go'i contains the most recent truth value.
  • go'e contains the second most recent truth value.
  • go'u contains the truth value of the first part of the most recent logical connective.
  • go'a contains the truth value of the second part of the most recent logical connective.

So if you want the program to execute ko brode if broda is true, and ko brodi if not, you could do this:

broda

.ijo ko brode

.i go'u ijonai ko brodi

By using the 'experimental cmavo' feature described later, you could reduce i go'u i jonai to i jo'ai for convenience.

Variables

The ko'a-series, la} followed by a name, or any sequence of lerfu can refer to a value.

Variables defined outside of a function are global. Variables can also be explicitly made global with the brivla vitno (as in ko'a vitno). To access a global variable inside a function, follow it with bi'u nai when it first appears. bi'u can also be used to explicitly state that a variable is local.

Assignment

Assignment of variables can be done with goi, dubgau, or dungau, which have different effects.

goi performs the action called "unification": if the variable on one side is assigned and the variable on the other side is not, it makes the unassigned variable equal to the assigned one. If both variables are already assigned, then the statement containing goi fails. If one variable is cleared with da'o, the other continues to exist.

Both variables are then given the same referent - this happens even if both variables were equal, or both variables were unassigned. So after unification, the variables are linked so that one changes when the other does. This can be used to give more useful names to items of a sequence, or to give a variable a lerfu-based name so it can be used in a mekso expression.

dungau takes three arguments:

dungau = x1 (ko) sets the value of x2 (variable) equal to that of x3 (variable)

The referent of the variable in x2 is not changed, so one can be assigned again later without affecting the other. This is the standard form of assignment in most programming languages.

dubgau takes any number of arguments after the first which is ko. Every argument starting at x2 is given the same referent as the rightmost argument.

Defining brivla

Brivla are defined using logically linked statements, with ca'e at the beginning of the bridi that is being defined. The connective should usually be o or onai, but a would work like onai because of the laziness rule. So a brivla could be defined using these general structures:

  • go ca'e <bridi being defined> gi <statement>
  • go ca'e <bridi being defined> gi ko zukte la'edi'e .i <block>
  • ko ca'e <bridi-tail being defined> gi'o <bridi-tail> (defines simple commands)
  • ca'e <bridi being defined> ijo <statement or block>

This last form is the one which is most likely to be useful for complex definitions.

Truth tests

If all the places are known and none are ko, the brivla is a truth test, which is true if the defining statement or block is true, and false otherwise.

Example

This brivla tests whether x1 is between the items in the two-item sequence x2.

ca'e ko'a jbini ko'e ijo (defining "x1 is between x2"):
tu'e
by goi le pamoi be ko'e (b = ko'e1)
.i cy goi le remoi be ko'e (c = ko'e2)
.i ge ko'a zmadu by gi ko'a mleca cy (is it true that b < ko'a < c?)
tu'u  

With this brivla defined:

.i li 3 jbini li 2 ce'o 4 (returns true)
.ijo ko cusku lu go'i li'u (prints "go'i")
.i go'i ijonai ko cusku lu na go'i li'u (otherwise print "na go'i")

Functions

A function is like a truth test, but one of its places (usually the x1) is unknown. The result is returned as that place, using the predicate jagdu'o.

jagdu'o = x1, as a result, is equal to x2

If x1 is unknown, jagdu'o sets x1 equal to x2. If x1 is known, jagdu'o does nothing, but has a truth value of whether x1 equals x2. This means that a function can be used as a truth test.

The result of a function is retrieved with le.

Example

This brivla can either return the square of x2 in x1, or test whether x1 equals x2 squared.

ca'e ko'a kurpi'i by ijo  
ko'a jagdu'o li by pi'i by  

With this brivla defined:

ko cusku le kurpi'i be li 3 (prints "9")
.i li 16 kurpi'i li 4 (returns true)
.ijo ko cusku lu go'i li'u (prints "go'i")

Commands

A command is a truth test or a function which performs an action. This could be accomplished simply with ko statements in the block, but it is good form to fix ko as the x1 by putting it in the x1 of the ca'e phrase. Hence, the definition of a command ganzu which takes one argument ko'a would look like this:

ca'e ko ganzu ko'a ijo
tu'e
...
tu'u

Then this command would be called with {ko ganzu ..., and if the ko were not there (for example, if something tried to refer to le ganzu, which makes little sense), the command would fail.

Loops

A "while" loop would be done in this form:

  • .i ca le du'u <statement1> kei <statement2>

...which says to repeatedly perform <statement2> while <statement1> is true. Of course, this could always be rephrased as

  • .i <statement2> ca le du'u <statement1>

Also, <statement2> would very frequently be a block.

Case statements

Case statements would be begun by asking the computer a question; it would then return a Lojban string containing the answer to that question. I'm not certain yet how the choices would be given from there.

Strings

Literal strings and filenames

Pure strings are quoted with zoi; filenames are quoted with la'o. The ordinary delimiters should be usable in most cases, though I suggest that the angle bracket convention (using < and > to delimit the quoted text) should be usable as well. The quoted string would be whatever is between the delimiters, not counting the whitespace on each end. These strings would be the only part of samtrosku that is not speakable.

If the string is a particularly pathological one, you may use a mekso of lerfu (example: "hi^Z ", where ^Z is character code 1A and the string ends with a space, could be lu'e me'o y'ybu ibu se'e pa dau na'a bu denpa bu.

Character codes should be hexadecimal. In C they are octal, but this is because the letters A-F would create ambiguity if they appeared in a character code embedded in a string. In samtrosku there is no such problem.

Obligatory troll-protective disclaimer: This does not mean that we endorse using hexadecimal "by default" in the rest of the language, either Lojban or samtrosku.

Lojban strings

Text enclosed in lo'u ... le'u is a string composed of Lojban words. The literal text is not used; the text is stored as the Lojban words it represents.

So ko cusku lo'u i ca le nu broda le'u could display .i ca le nu broda.

Input and output

Besides the fact that cusku would write to standard output, I haven't yet decided anything here.

Loading brivla definitions

The definition of a brivla can be loaded from another file.

A brivla is defined using cilre.

cilre = x1 (ko) learns the instructions corresponding to x3 (gismu/lujvo/fu'ivla) from x4 (file)

x2 and x5 are ignored, though they might mean something in a later revision of samtrosku. If x3 is zo'e it learns every brivla in that file; if x4 is zo'e an implementation may search a standard path to find the appropriate brivla.

Examples:

  • ko cilre fi zo cunyna'u
  • ko cilre fi zo pandi la'o < /usr/share/samtrosku/brivla/vrici.sam >
  • ko cilre fo la'o < /usr/share/samtrosku/brivla/kanji.sam >

Defining experimental cmavo

A cmavo is defined in experimental cmavo space by setting it equal to a fragment of text, using the predicate cmavygau. {gl|cmavygau|x1 (ko) defines x2 (cmavo) to represent x4 (text fragment)}} x3 (the selma'o) and x5 (samtrosku) are ignored.

For example:

.i ko cmavygau zo lai'e fi makes "la'edi'e" shorter
lo'u la'e di'e le'u
.i ko cmavygau zo jo'ai fi makes a pseudo-connective for "if-then-else"
lo'u go'u ijonai le'u

sumti types

Samtrosku is weakly typed, and inherently to the semantics of Lojban, there are lots of types. Only certain types work in certain situations. Here is a list of types and examples of how each might be referred to (these references could then be assigned to a variable such as ko'a using goi.)

For a better list of types, look through the gismu list. Some places are specified for ka, others for du'u, some for sets and others for numbers. These are some Lojban datatypes.