See the Text_Editor page for details of the editor and the usage of the text syntax.
Screenshots which give an impression of the syntax in the text editor are shown on the screenshot page.
EBNF for the TextEditor's syntax
parse_unit = 'machine' identifier_literal comment* ('refines' identifier_literal (',' identifier_literal)*)? ('sees' identifier_literal (',' identifier_literal)*)? ('variables' identifier_list)? ('invariants' labeled_predicate_with_theorem+)? ('variant' variant)? ('events' event+)? end | 'context' identifier_literal comment* ('extends' identifier_literal (',' identifier_literal)*)? ('sets' identifier_list)? ('constants' identifier_list)? ('axioms' labeled_predicate_with_theorem+)? end ; variant = expression comment*; event = convergence? 'event' identifier_literal comment* event_refinement? ('any' identifier_list)? ('where' labeled_predicate+)? ('with' labeled_predicate+)? ('then' labeled_action+)? ; convergence = 'ordinary' | 'convergent' | 'anticipated' ; event_refinement = 'refines' identifier_literal (',' identifier_literal)* | 'extends' identifier_literal ; labeled_action = '@' label action comment*; labeled_predicate = '@' label predicate comment*; labeled_predicate_with_theorem = '@' label predicate comment*| 'theorem' '@' identifier_literal predicate comment*; identifier = identifier_literal comment*; identifier_list = identifier identifier* ; comment = '//' string-to-next-linebreak | '/*' string '*/' ;
NOTE: the EBNF above for Contexts specifies that constants come after sets. It seems to be the other way around in RODIN.
As you might notice the grammar only describes the structure of machines and contexts. For the terminals predicate, expression and action you can use the syntax for formulas which you already know from editing models with the graphical Event-B editor. You can use the ASCII representation of mathematical symbols or their unicode. See the Rodin help for details about the mathematical symbols and their ASCII counterparts.
Also note that the rather strict position for comments is necessary. In the RodinDB most elements can be commented. That means if you comment an element this comment is attached to the element. To reflect this in a text syntax we need to exactly specify to which textual element a comment belongs. Therefore we decided to always associate a comment to the previous element. Additionally database only allows one comment per element. The grammar allows multiple comments, but to reflect the database restriction to one internally multiple comments are concatenated to one with linebreaks between the separate comments.
The syntax doesn't allow whitespaces within labels. It assumes that all characters following an '@' belong to the label until the next whitespace character is found.
Feel free to add own examples!
Machine 'm0' from example 'maximum'
machine m0 sees array variables i invariants @inv1 i ∈ ℕ events event INITIALISATION then @act1 i :∈ ℕ end event find_max any j where @grd1 j ∈ 1‥n @grd2 ∀k·k∈1‥n ⇒ a(k) ≤ a(j) then @act1 i ≔ j end end
Context 'array' from example 'maximum'
context array constants n a axioms @axm1 n ∈ ℕ1 @axm2 a ∈ 1‥n → ℤ end