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
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176

//! The `dnm::parameters` submodule provides data structures for customizing
//! and configuring a DNM's construction and use

use libxml::readonly::RoNode;
use std::collections::HashMap;
use std::fmt;
use std::sync::Arc;

/// Some temporary data for the parser
#[derive(Debug)]
pub struct RuntimeParseData {
  /// plaintext is currently terminated by some whitespace
  pub had_whitespace: bool,
  /// plaintext representation as vector of chars (to deal with UTF-8 mess)
  /// TODO: Use plaintext/byte_offsets directly instead
  pub chars: Vec<char>,
}
impl Default for RuntimeParseData {
  fn default() -> RuntimeParseData {
    RuntimeParseData {
      had_whitespace: true, // skip leading whitespace
      chars: Vec::new(),
    }
  }
}

/// Specifies how to deal with a certain tag
#[derive(Clone)]
pub enum SpecialTagsOption {
  /// Recurse into tag (default behaviour)
  Enter,
  /// Normalize tag, replacing it by some token
  Normalize(String),
  /// Normalize tag, obtain replacement string by function call
  FunctionNormalize(Arc<fn(RoNode) -> String>),
  /// Skip tag
  Skip,
}

impl fmt::Debug for SpecialTagsOption {
  fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
    use SpecialTagsOption::*;
    write!(f, "SpecialTagsOption {{")?;
    match self {
      Enter => write!(f, "Enter")?,
      Skip => write!(f, "Skip")?,
      Normalize(v) => write!(f, "Normalize({v})")?,
      FunctionNormalize(_) => write!(f, "FunctionNormalize")?,
    };
    write!(f, "}}")
  }
}

/// Parameters for the DNM generation
#[derive(Debug, Clone)]
pub struct DNMParameters {
  /// How to deal with special tags (e.g. `<math>` tags)
  pub special_tag_name_options: HashMap<String, SpecialTagsOption>,
  /// How to deal with tags with special class names (e.g. ltx_note_mark)
  /// *Remark*: If both a tag name and a tag class match, the tag name rule
  /// will be applied.
  pub special_tag_class_options: HashMap<String, SpecialTagsOption>,
  /// merge sequences of whitespaces into a single ' '.
  /// *Doesn't affect tokens*
  pub normalize_white_spaces: bool,
  /// put spaces before and after tokens
  pub wrap_tokens: bool,
  /// Replace unicode characters by the ascii code representation
  pub normalize_unicode: bool,
  /// Apply the morpha stemmer once to the text nodes
  pub stem_words_once: bool,
  /// Apply the morpha stemmer to the text nodes
  /// as often as it changes something
  pub stem_words_full: bool,
  /// Move to lowercase (remark: The stemmer does this automatically)
  pub convert_to_lowercase: bool,
  /// Support back mapping, i.e. mapping plaintext offsets back to the DOM
  pub support_back_mapping: bool,
}

impl Default for DNMParameters {
  /// Don't do anything fancy and specific by default
  fn default() -> DNMParameters {
    DNMParameters {
      special_tag_name_options: HashMap::new(),
      special_tag_class_options: HashMap::new(),
      normalize_white_spaces: true,
      wrap_tokens: false,
      normalize_unicode: false,
      stem_words_once: false,
      stem_words_full: false,
      convert_to_lowercase: false,
      support_back_mapping: true,
    }
  }
}

impl DNMParameters {
  /// Normalize in a reasonable way for our math documents
  pub fn llamapun_normalization() -> DNMParameters {
    let mut name_options = HashMap::new();
    name_options.insert(
      "math".to_string(),
      SpecialTagsOption::Normalize("mathformula".to_string()),
    );
    name_options.insert(
      "cite".to_string(),
      SpecialTagsOption::Normalize("CitationElement".to_string()),
    );
    name_options.insert("img".to_string(), SpecialTagsOption::Skip);
    name_options.insert("table".to_string(), SpecialTagsOption::Skip);
    name_options.insert("head".to_string(), SpecialTagsOption::Skip);
    name_options.insert("footer".to_string(), SpecialTagsOption::Skip);

    let mut class_options = HashMap::new();
    class_options.insert(
      "ltx_equation".to_string(),
      SpecialTagsOption::Normalize("\nmathformula\n".to_string()),
    );
    class_options.insert(
      "ltx_equationgroup".to_string(),
      SpecialTagsOption::Normalize("\nmathformula\n".to_string()),
    );
    class_options.insert(
      "ltx_ref".to_string(),
      SpecialTagsOption::Normalize("REF".to_string()),
    );
    class_options.insert("ltx_authors".to_string(), SpecialTagsOption::Skip);
    class_options.insert("ltx_TOC".to_string(), SpecialTagsOption::Skip);
    class_options.insert("ltx_note_mark".to_string(), SpecialTagsOption::Skip);
    class_options.insert("ltx_note_outer".to_string(), SpecialTagsOption::Skip);
    class_options.insert("ltx_bibliography".to_string(), SpecialTagsOption::Skip);
    // Ignores all caption metadata tags, to avoid leaking artefacts into a pure language target
    // TODO: Is there merit to extending this to ignoring all ltx_tag elements? leaving things as-is
    // allows for some curious artefacts to sneak into the plain-text files, such as
    // bullets/numbers from \item commands
    class_options.insert("ltx_tag_figure".to_string(), SpecialTagsOption::Skip);
    class_options.insert("ltx_tag_table".to_string(), SpecialTagsOption::Skip);

    DNMParameters {
      special_tag_name_options: name_options,
      special_tag_class_options: class_options,
      normalize_white_spaces: false, /* Keeping it raw for tokenization best results, newlines
                                      * are meaningful */
      /* important for cases where we have things like x$\prime\prime$, and risk creating a single word "xmathformula" instead of the lexical "x mathformula"
        There may be better tokenization tricks to employ later on (in the word tokenizer), but for now wrapping seems necessary
      */
      wrap_tokens: true,
      normalize_unicode: true,
      ..Default::default()
    }
  }

  /// Prints warnings, if the parameter settings don't make sense.
  /// Doesn't check for every possible stupidity
  pub fn check(&self) {
    if self.stem_words_once && self.stem_words_full {
      dbg!(
        "llamapun::dnm: Parameter options stem_words_once\
         and stem_words_full are both set"
      );
    }
    if (self.stem_words_once || self.stem_words_full) && self.convert_to_lowercase {
      dbg!(
        "llamapun::dnm: Parameter option convert_to_lowercase\
         is redundant, because stemming converts to lowercase already"
      );
    }
    if (self.stem_words_once || self.stem_words_full) && self.support_back_mapping {
      dbg!(
        "llamapun::dnm: Parameter option support_back_mapping\
         does not work in combination with word stemming yet"
      );
    }
  }
}