Getting started

Narsese grammar

This is the Narsese syntax and grammar used by the nars.dll library. It was adapted from Wang’s NAL Specification documentation.

                         ⟨sentence⟩ ::=  ⟨judgment⟩ | ⟨goal⟩ | ⟨question⟩
                         ⟨judgment⟩ ::=  [⟨tense⟩] ⟨statement⟩. ⟨truth-value⟩
                             ⟨goal⟩ ::=  ⟨statement⟩! ⟨desire-value⟩
                         ⟨question⟩ ::=  (⟨term⟩ ⟨copula⟩ ⟨term⟩) | ⟨term⟩
		                      | (¬ ⟨statement⟩)
                                      | (∧ ⟨statement⟩ ⟨statement⟩⁺)
                                      | (∨ ⟨statement⟩ ⟨statement⟩⁺)
                                      | (∧/ ⟨statement⟩ ⟨statement⟩⁺)
                                      | (∧| ⟨statement⟩ ⟨statement⟩⁺)
                                      | (⇑ ⟨word⟩ ⟨term⟩*)
                          ⟨copula⟩ ::=   → | ↔ | ∘→ | →∘ | ∘→∘  
                                      | ⇒ | ⇔ | /⇒ |  \⇒ | |⇒ | /⇔ | |⇔
                           ⟨tense⟩ ::=   /⇒ |  \⇒ | |⇒
                            ⟨term⟩ ::=   ⟨word⟩ | ⟨variable⟩ | ⟨statement⟩
                                      | {⟨term⟩⁺} | [⟨term⟩⁺]
                                      | (∩ ⟨term⟩ ⟨term⟩⁺)
                                      | (∪ ⟨term⟩ ⟨term⟩⁺)
                                      | (- ⟨term⟩ ⟨term⟩)
                                      | (⊖ ⟨term⟩ ⟨term⟩)
                                      | (× ⟨term⟩ ⟨term⟩⁺)
                                      | (⊥ ⟨term⟩* ◊ ⟨term⟩*)
                                      | (⊤ ⟨term⟩* ◊ ⟨term⟩*)
                       ⟨variable⟩ ::=    ⟨independent-variable⟩
                                      | ⟨dependent-variable⟩
                                      | ⟨query-variable⟩
           ⟨independent-variable⟩ ::=    #⟨word⟩
             ⟨dependent-variable⟩ ::=    #[⟨word⟩(⟨independent-variable⟩*)]
                 ⟨query-variable⟩ ::=    ?[⟨word⟩]
                    ⟨truth-value⟩ ::=    <⟨real-number⟩, ⟨real-number⟩>
                   ⟨desire-value⟩ ::=    <⟨real-number⟩, ⟨real-number⟩>
                           ⟨word⟩ ::=    [_a-zA-Z0-9ɐ-ʯæ₀-₉]⁺

Special symbols of Narsese are not replaced by a combination of ASCII character but rather are encoded with UTF-8. Narsese code can be edited by any editor supporting UTF-8 encoding. NarsDevelop, provided on this website, is a simple GUI used togehter with the NARS library, which contains a simple editor of Narsese code with possibility to directly input special symbols used in Narsese by a keyboard shortcut. It has also syntaxhilighting and some debuging capability of Narsese code.

In case, compounds contain exactly two terms, operators ∧, ∨, ∩, ∪, -, ⊖, × can use both - prefix and infix form (e.g.: (∧, S1, S2) and (S1 ∧ s2) are equivalent).

Control and debug commands

The following is an extension of the grammar, used just during development and debugging and it is not part of the Narsese language.

                 space (ignored) ::= /\s\n\r\t+/
     multiline-comment (ignored) ::= '/*' .. '*/'
    singleline-comment (ignored) ::= /\#[^\n\r]*/

                          ⟨line⟩  ::=   ⟨sentence⟩ | ⟨control⟩
                        ⟨control⟩ ::=   ⋙ ⟨commands⟩
                       ⟨commands⟩ ::=   clear
                                      | sync						
                                      | test ⟨sentence⟩
                                      | words
                                      | compounds
                                      | statements
                                      | memory
                                      | question
                                      | inference
                                      | debug on | debug off
                                      | version

For performance reason, the answer from the system for these commands takes a mixture of text and binary form.

nars library uses C++ style single and multi-line comments.

Everything followed by a special symbol ⋙ are processed as a command.

• clear – clears system memory. Useful, if we want to test some test cases or examples, otherwise, previous experience of the system might affect results.
• sync – waits till all inference threads are in waiting state. This is useful only for the test command (which implicitly uses it). All tasks run concurrently, and might happen, that incidentally a hint would be given before conclusion is found. It might be.
• test ⟨sentence⟩ - test the last outcome from the system against the expected one (sentence) in terms of syntax and truth value.
• debug on | debug off – with debug on, we get some additional information about the inference process. Some commands requires to be debug set on.
• words – gets words from the memory.

  Output takes a mix of text and binary form:

  Length [Bytes]	3	4	termSize1	n	1	...
  Content	"<w>"	termSize	termId02	word03	'\0'4	...

  ¹termSize is a size of termId - internal representation of a term. termSize currently is represented by 8 bytes (64 bits).
  ²A binary representation (identifier) of a term.
  ³A text (string) representation of a term.
  ⁴Null termination of a string.

• compounds – gets compound terms from the memory.

  Length [Bytes]	3	4	termSize	n	1	...
  Content	"<c>"	termSize	termId0	compound0	'\0'	...

• statements – gets statements from the memory.

  Length [Bytes]	3	4	termSize	n	1	...
  Content	"<s>"	termSize	termId0	statement0	'\0'	...

• memory – print system memory in pure Narsese. Similar to 'statements', but every type of the same statement is printed separately. The output from this command can be used as input to the system for memory reloading.

  Length [Bytes]	3	n	1	...
  Content	"<m>"	statement0	'\n'	...

• question – information about question derivation of the last question entering to the system. To able to use this command, debug must be on. If the question can’t be solved by any inference process, sub questions are generated during inference process which leads to the original question. This process is done recursively, until at least one question can be solved. Then tracking back we get the solution for the original question.

  Header

  Length [Bytes]	3	4	4	4	4
  Content	"<q>"	termSize	numberOfEdges	numberOfNodes	orderSize

  A list of edges

  Length [Bytes]	termSize	termSize	...
  Content	node0	node1	...

  A list of nodes

  Length [Bytes]	termSize	termSize	1	...
  Content	termId	statement	'\0'	...

  Processing sequence (order of processing sub questions)

  Length [Bytes]	termSize	termSize	...
  Content	termId	attribute	...

  NarsDevelop transforms these information into an SVG graph wich can be saved into a computer storage.

• inference – Shows inference processes leading from a question to the final conclusion.

  Header

  Length [Bytes]	3	4	4	4
  Content	"<n>"	statementsOffset	statementsNum	opcodeNum

  Inference

  Length [Bytes]	4	termSize	termSize	...	4	...
  Content	premiseNum	premise0	premise1	...	opcode	...

  Statements

  Length [Bytes]	termSize	n	1	...
  Content	termId	statement0	'\0'	...

  Operation Code

  Length [Bytes]	4	n0	1	n1	1	...
  Content	OpCode	description0	'\0'	description1	'\0'	...

• version – prints out the version of the nars library in form of a string.

  Length [Bytes]	3	n	1
  Content	<v>	library version	'\0'

• Other outputs:
  • Output from internal questions

    Length [Bytes]	3	n	1
    Content	"<i>"	Narsese statement	'\0'

  • Time and elapsed cycles for the last external question.

    Length [Bytes]	3	4	4
    Content	"<t>"	Elapsed time [μs]	number of inference cycles

  • Error reporting.

    Length [Bytes]	3	4	4	4	n	1
    Content	"<e>"	row	column	error type1	description	'\0'

  ¹Possible type of error:

  • 0x1 - syntax error
  • 0x2 - different answer from expected one (applies, when the test command is used)
  • 0x4 - different truth value (same as above)
  • 0x8 - different tense (same as above)

Direct input of special symbols

NarsDevelop has some special shortcuts to directly enter special symbols of Narsese using keyboard as it is shown in the following table.

• Prev
• Next