Struct SectionDetectionV2Config 

pub struct SectionDetectionV2Config {
    pub line_height_tolerance: f32,
    pub font_size_tolerance: f32,
    pub structural_size_margin: f32,
    pub structural_size_ratio: Option<f32>,
    pub min_alpha_ratio: f32,
    pub max_depth: u32,
    pub enforce_max_depth: bool,
    pub starting_section_level: u32,
    pub inclusion_patterns: Vec<InclusionPattern>,
    pub inclusion_max_length: usize,
    pub exclusion_patterns: Vec<String>,
    pub tiebreaker_keywords: Vec<TiebreakerKeyword>,
}

Configuration for the V2 section detection rule (V3 algorithm — Block 09).

Three-tier piecewise classifier on delta = font_size - body_size plus pre-gates (rotation, alpha-ratio) and pattern refinement (inclusion / exclusion regex) as a backup. Isolation is leaf-based, consulting the Placement.region_label set by analytics::reading_order::tag_and_resort.

Fields

line_height_tolerance: f32

Y-coordinate tolerance (points) for grouping bboxes onto the same visual line. Two elements within |Δy| < this in the same Region tree leaf are considered to be on the same baseline. Defaults to a value smaller than typical inter-line spacing so consecutive lines do not merge.

font_size_tolerance: f32

Font-size tolerance (points). Defines the symmetric ±tolerance band around body size.

• delta < -tolerance → REJECT (below-body noise).
• |delta| ≤ tolerance → R3 (at-body band): needs bold AND isolated_in_leaf.
• tolerance < delta ≤ structural_size_margin → R2 (medium): needs bold OR isolated_in_leaf.
• delta > structural_size_margin → R1 (large): auto-promote unconditionally.

structural_size_margin: f32

Size margin (points) above body text at which size alone confirms structural role. Region 1 threshold: delta > structural_size_margin → auto-promote.

structural_size_ratio: Option<f32>

Proportional alternative to structural_size_margin. When Some, Region 1 threshold is body_size * ratio instead of body_size + margin. Default None (use margin).

min_alpha_ratio: f32

Minimum alphabetic character ratio for a candidate to survive (inherits semantics from old rule’s min_alpha_ratio).

max_depth: u32

Max hierarchy depth (inherits from old rule).

enforce_max_depth: bool

starting_section_level: u32

inclusion_patterns: Vec<InclusionPattern>

Regex patterns that promote a weak/rejected candidate to a section (escape hatch — e.g., “^\d+\.\d+” for numbered subsections). Promotion additionally requires the per-pattern structural gates (require_bold, require_isolation) and a global length cap (inclusion_max_length). See CR-26 (length cap, isolation) and CR-42 (per-pattern bold/isolation gating).

inclusion_max_length: usize

Maximum text length (in characters) for an inclusion-pattern match to promote. Real structural labels (“Article 64”, “CHAPTER II”) are short; body wrap-lines that happen to begin with a structural keyword are long. This is the synthetic gate Pass 2 needs because, unlike Pass 1, it has no bold/rarity confirming signal — pattern + isolation alone admit recital wrap-lines on documents like CELEX where font_size is degenerate.

exclusion_patterns: Vec<String>

Regex patterns that demote a promoted candidate back to non-section (escape hatch — e.g., “^Figure\s” for figure captions).

tiebreaker_keywords: Vec<TiebreakerKeyword>

Ordered list of (keyword_name, regex) pairs that identify the structural “tier” of a section. Consulted only when the font-size delta vs. the previous section is within font_size_tolerance (the tie). When the tie fires, keyword identity decides whether the new section is a sibling, a step-back-up to an earlier tier, or a deeper tier.

Order matters: the first matching pattern wins. Place specific keywords before generic ones (e.g. structural words before bare-numbered).

Trait Implementations

impl Clone for SectionDetectionV2Config

fn clone(&self) -> SectionDetectionV2Config

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for SectionDetectionV2Config

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.

impl Default for SectionDetectionV2Config

fn default() -> SectionDetectionV2Config

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for SectionDetectionV2Config

fn deserialize<__D>( __deserializer: __D, ) -> Result<SectionDetectionV2Config, <__D as Deserializer<'de>>::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for SectionDetectionV2Config

fn serialize<__S>( &self, __serializer: __S, ) -> Result<<__S as Serializer>::Ok, <__S as Serializer>::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.