loftspec 2.0.2

The Loft Specification Format: document model, identity, canonical bytes, hashing and conversion
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141

//! Recognising the five basic EARS forms over a sentence's tokens: the same classification as
//! the language rule that reports a statement outside the five forms, so a sentence the Format
//! stores as slots is one that rule accepts. Spans index [`crate::text::tokenize`].

use std::ops::Range;

use crate::text::lowercase;

/// One of the five basic EARS forms, without its slots.
#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
pub enum Form {
    Ubiquitous,
    EventDriven,
    StateDriven,
    Unwanted,
    Optional,
}

/// A half-open `[start, end)` range of token indexes.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub(crate) struct Span {
    pub(crate) start: usize,
    pub(crate) end: usize,
}

/// A sentence split into its precondition, system and response spans. The ubiquitous form has
/// no precondition. The system span leaves out a leading `the`.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub(crate) struct Parse {
    pub(crate) form: Form,
    pub(crate) precondition: Option<Span>,
    pub(crate) response: Span,
}

/// Parses `text`, whose tokens lie at `ranges` (as [`crate::text::token_ranges`] gives them), as
/// one anchored basic EARS form; combined forms and empty placeholders give `None`. The system
/// phrase may be a common noun with `the` or a bare proper name: it starts after the comma that
/// closes the precondition, or failing that at the precondition's `the`.
pub(crate) fn parse(text: &str, ranges: &[Range<usize>]) -> Option<Parse> {
    let tokens: Vec<String> = ranges
        .iter()
        .map(|range| lowercase(&text[range.clone()]))
        .collect();
    let comma_before =
        |index: usize| index > 0 && text[ranges[index - 1].end..ranges[index].start].contains(',');
    ubiquitous(&tokens, &comma_before)
        .or_else(|| prefixed(&tokens, &comma_before, "when", Form::EventDriven))
        .or_else(|| prefixed(&tokens, &comma_before, "while", Form::StateDriven))
        .or_else(|| unwanted(&tokens))
        .or_else(|| prefixed(&tokens, &comma_before, "where", Form::Optional))
}

/// A ubiquitous system phrase is a whole noun phrase: it opens with no prefix word or `then`,
/// holds no comma, and a bare proper name holds no later `the` that would mark a lead-in clause.
fn ubiquitous(tokens: &[String], comma_before: &dyn Fn(usize) -> bool) -> Option<Parse> {
    let first = tokens.first()?;
    if is_prefix(first) || first == "then" {
        return None;
    }
    let shall = shall_after(tokens, 1)?;
    if (1..=shall).any(comma_before)
        || (first != "the" && tokens[1..shall].iter().any(|token| token == "the"))
    {
        return None;
    }
    system_is_nonempty(tokens, 0, shall)?;
    Some(Parse {
        form: Form::Ubiquitous,
        precondition: None,
        response: span(shall + 1, tokens.len()),
    })
}

fn prefixed(
    tokens: &[String],
    comma_before: &dyn Fn(usize) -> bool,
    prefix: &str,
    form: Form,
) -> Option<Parse> {
    if tokens.first().is_none_or(|token| token != prefix) {
        return None;
    }
    let shall = shall_after(tokens, 2)?;
    let comma = (2..shall).rev().find(|&index| comma_before(index));
    let boundary = comma.or_else(|| (2..shall).find(|&index| tokens[index] == "the"))?;
    if contains_other_prefix(&tokens[1..boundary], prefix) {
        return None;
    }
    system_is_nonempty(tokens, boundary, shall)?;
    Some(Parse {
        form,
        precondition: Some(span(1, boundary)),
        response: span(shall + 1, tokens.len()),
    })
}

fn unwanted(tokens: &[String]) -> Option<Parse> {
    if tokens.first().is_none_or(|token| token != "if") {
        return None;
    }
    (2..tokens.len().saturating_sub(1)).find_map(|then| {
        if tokens[then] != "then" || contains_other_prefix(&tokens[1..then], "if") {
            return None;
        }
        let shall = shall_after(tokens, then + 2)?;
        system_is_nonempty(tokens, then + 1, shall)?;
        Some(Parse {
            form: Form::Unwanted,
            precondition: Some(span(1, then)),
            response: span(shall + 1, tokens.len()),
        })
    })
}

/// The system phrase in `[start, shall)`, without a leading `the`, is not empty.
fn system_is_nonempty(tokens: &[String], start: usize, shall: usize) -> Option<()> {
    let start = if tokens.get(start).is_some_and(|token| token == "the") {
        start + 1
    } else {
        start
    };
    (start < shall).then_some(())
}

fn is_prefix(token: &str) -> bool {
    matches!(token, "when" | "while" | "if" | "where")
}

const fn span(start: usize, end: usize) -> Span {
    Span { start, end }
}

fn shall_after(tokens: &[String], first_position: usize) -> Option<usize> {
    (first_position..tokens.len().saturating_sub(1)).find(|&index| tokens[index] == "shall")
}

fn contains_other_prefix(tokens: &[String], allowed: &str) -> bool {
    tokens
        .iter()
        .any(|token| is_prefix(token) && token != allowed)
}