dsntk-recognizer 0.3.0

//! # Decision table builder

use crate::errors::*;
use crate::recognizer::Recognizer;
use crate::utils::get_allowed_and_default_output_values;
use crate::{AnnotationClause, AnnotationEntry, DecisionRule, DecisionTable, HitPolicy, InputClause, InputEntry, OutputClause, OutputEntry};
use dsntk_common::Result;

/// Combines sizes of individual components of the decision table.
struct Size {
  input_clauses_count: usize,
  input_values_count: usize,
  output_clauses_count: usize,
  output_components_count: usize,
  output_values_count: usize,
  annotation_clauses_count: usize,
  rule_count: usize,
}

/// Validates the size of the individual components of the decision table in relation to each other.
fn validate_size(recognizer: &Recognizer) -> Result<Size> {
  // decision table may have zero or more input clauses
  let input_clauses_count = recognizer.input_clause_count;
  // number of input expressions must be equal to the number of input clauses
  let input_expression_count = recognizer.input_expressions.len();
  if input_expression_count != input_clauses_count {
    return Err(err_invalid_size(&format!(
      "number of input expressions ({input_expression_count}) must be equal to the number of input clauses ({input_clauses_count})"
    )));
  }
  // when input values are present, then the number of input values must be equal to the number input expressions
  let input_values_count = recognizer.allowed_input_values.len();
  if input_values_count > 0 && input_values_count != input_clauses_count {
    return Err(err_invalid_size(&format!(
      "number of input values ({input_values_count}) must be equal to the number of input clauses ({input_clauses_count})"
    )));
  }
  // decision table must have minimum one output clause
  let output_clauses_count = recognizer.output_clause_count;
  if output_clauses_count == 0 {
    return Err(err_invalid_size("decision table must have minimum one output clause"));
  }
  // the number of the output components must be equal to the number of output clauses when more than one output clause is present
  let output_components_count = recognizer.output_components.len();
  if output_clauses_count > 1 {
    if output_components_count != output_clauses_count {
      return Err(err_invalid_size(&format!(
        "number of output components ({output_components_count}) must be equal to the number of output clauses ({output_clauses_count})"
      )));
    }
  } else if output_components_count != 0 {
    return Err(err_invalid_size("number of output components must be zero"));
  }
  // when output values are present, then the number of output values must be equal to the number of output clauses
  let output_values_count = recognizer.allowed_output_values.len();
  if output_values_count > 0 && output_values_count != output_clauses_count {
    return Err(err_invalid_size(&format!(
      "number of output values ({output_values_count}) must be equal to the number of output clauses ({output_clauses_count})"
    )));
  }
  // decision table must contain minimum one rule
  let rule_count = recognizer.rule_count;
  if rule_count == 0 {
    return Err(err_invalid_size("decision table must contain minimum one rule"));
  }
  // number of rows of input entries must be equal to the number of rules
  let input_entries_row_count = recognizer.input_entries.len();
  if input_entries_row_count != rule_count {
    return Err(err_invalid_size(&format!(
      "number of input entries ({input_entries_row_count}) must be equal to the number of rules ({rule_count})",
    )));
  }
  // number of input entries in each row must be equal to the number of input clauses
  for (row_index, row) in recognizer.input_entries.iter().enumerate() {
    if row.len() != input_clauses_count {
      return Err(err_invalid_size(&format!(
        "number of input entries ({}) must be equal to the number of input clauses ({input_clauses_count}) in row {row_index}",
        row.len(),
      )));
    }
  }
  // number of rows of output entries must be equal to the number of rules
  let output_entries_row_count = recognizer.output_entries.len();
  if output_entries_row_count != rule_count {
    return Err(err_invalid_size(&format!(
      "number of output entries ({output_entries_row_count}) must be equal to the number of rules ({rule_count})"
    )));
  }
  // number of output entries in each row must be equal to the number of output clauses
  for (row_index, row) in recognizer.output_entries.iter().enumerate() {
    if row.len() != output_clauses_count {
      return Err(err_invalid_size(&format!(
        "number of output entries ({}) must be equal to the number of output clauses ({output_clauses_count}) in row {row_index}",
        row.len()
      )));
    }
  }
  // decision table may contain some annotations
  let annotation_clauses_count = recognizer.annotation_clause_count;
  if annotation_clauses_count > 0 {
    // number of rows of annotation entries must be equal to the number of rules
    let annotation_entries_row_count = recognizer.annotation_entries.len();
    if annotation_entries_row_count != rule_count {
      return Err(err_invalid_size(&format!(
        "number of annotation entries ({annotation_entries_row_count}) must be equal to the number of rules ({rule_count})"
      )));
    }
    // number of annotation entries in each row must be equal to the number of annotation clauses
    for (row_index, row) in recognizer.annotation_entries.iter().enumerate() {
      if row.len() != annotation_clauses_count {
        return Err(err_invalid_size(&format!(
          "number of annotation entries ({}) must be equal to the number of annotation clauses ({}) in row {}",
          row.len(),
          annotation_clauses_count,
          row_index
        )));
      }
    }
  }
  Ok(Size {
    input_clauses_count,
    input_values_count,
    output_clauses_count,
    output_components_count,
    output_values_count,
    annotation_clauses_count,
    rule_count,
  })
}

/// Recognizes a decision table defined as Unicode text.
pub fn from_unicode(text: &str, trace: bool) -> Result<DecisionTable> {
  // Recognize the decision table.
  let recognizer = Recognizer::recognize(text, trace)?;

  // Validate the size of the individual parts of the decision table.
  let size = validate_size(&recognizer)?;

  // Prepare the information item name.
  let information_item_name = recognizer.information_item_name.clone();

  // Prepare the hit policy.
  let hit_policy = recognizer.hit_policy;

  // Prepare the aggregation.
  let aggregation = if let HitPolicy::Collect(built_in_aggregator) = hit_policy {
    Some(built_in_aggregator)
  } else {
    None
  };

  // Prepare the preferred orientation.
  let preferred_orientation = recognizer.orientation;

  // Prepare the output label
  let output_label = recognizer.output_label.clone();

  // Prepare the input clauses.
  let mut input_clauses = vec![];
  for i in 0..size.input_clauses_count {
    input_clauses.push(InputClause {
      input_expression: recognizer.input_expressions[i].clone(),
      allowed_input_values: if size.input_values_count > 0 {
        recognizer.allowed_input_values[i].clone()
      } else {
        None
      },
    });
  }

  // Prepare output clauses.
  let mut output_clauses = vec![];
  for i in 0..size.output_clauses_count {
    let output_component_name = if size.output_components_count > 0 {
      recognizer.output_components[i].clone()
    } else {
      output_label.clone()
    };
    let (allowed_output_values, default_output_value) = get_allowed_and_default_output_values(&if size.output_values_count > 0 {
      recognizer.allowed_output_values[i].clone()
    } else {
      None
    });
    output_clauses.push(OutputClause {
      output_component_name,
      allowed_output_values,
      default_output_value,
    });
  }

  // Prepare annotations.
  let mut annotations = vec![];
  for i in 0..recognizer.annotation_clause_count {
    annotations.push(AnnotationClause {
      name: recognizer.annotations[i].clone(),
    });
  }

  // Prepare rules.
  let mut rules = vec![];
  for rule_index in 0..size.rule_count {
    let mut input_entries = vec![];
    for column_index in 0..size.input_clauses_count {
      let input_entry = InputEntry {
        text: recognizer.input_entries[rule_index][column_index].clone(),
      };
      input_entries.push(input_entry);
    }
    let mut output_entries = vec![];
    for column_index in 0..size.output_clauses_count {
      let output_entry = OutputEntry {
        text: recognizer.output_entries[rule_index][column_index].clone(),
      };
      output_entries.push(output_entry);
    }
    let mut annotation_entries = vec![];
    for column_index in 0..size.annotation_clauses_count {
      let annotation_entry = AnnotationEntry {
        text: recognizer.annotation_entries[rule_index][column_index].clone(),
      };
      annotation_entries.push(annotation_entry);
    }
    rules.push(DecisionRule {
      input_entries,
      output_entries,
      annotation_entries,
    });
  }

  // Return the recognized decision table.
  Ok(DecisionTable::new(
    information_item_name,
    input_clauses,
    output_clauses,
    annotations,
    rules,
    hit_policy,
    aggregation,
    preferred_orientation,
    output_label,
  ))
}