Struct TableDetectionConfig 

pub struct TableDetectionConfig {
    pub enabled: bool,
    pub horizontal_strategy: TableStrategy,
    pub vertical_strategy: TableStrategy,
    pub column_tolerance: f32,
    pub row_tolerance: f32,
    pub min_table_cells: usize,
    pub min_table_columns: usize,
    pub regular_row_ratio: f32,
    pub max_table_columns: usize,
    pub column_merge_threshold: f32,
    pub v_split_gap: f32,
    pub text_fallback: bool,
}

Configuration for spatial table detection.

Fields

enabled: bool

Whether table detection is enabled.

horizontal_strategy: TableStrategy

Strategy for horizontal boundary detection.

vertical_strategy: TableStrategy

Strategy for vertical boundary detection.

column_tolerance: f32

X-coordinate tolerance for column grouping.

row_tolerance: f32

Y-coordinate tolerance for row grouping.

min_table_cells: usize

Minimum number of cells required for a valid table.

min_table_columns: usize

Minimum number of columns required for a valid table.

regular_row_ratio: f32

Ratio of regular rows required for a valid table structure.

max_table_columns: usize

Maximum number of columns allowed before rejecting as false positive.

column_merge_threshold: f32

Merge threshold for post-clustering column merge pass. Adjacent columns whose centers are within this distance are merged.

v_split_gap: f32

Minimum gap between Y-range groups of vertical lines to trigger a cluster split. Default: 20.0. Use smaller values (e.g. 4.0) for strict mode, larger (e.g. 40.0) for relaxed mode where V-lines at mixed Y-ranges should stay together.

text_fallback: bool

Enable text-only spatial detection as a fallback when no ruling lines are found.

When true and the page has no table-relevant paths (no ruling lines or rectangles), the detector falls through to detect_tables_from_spans_column_aware rather than returning an empty result. This is the right default for structured output callers (to_markdown, to_html) that explicitly want tabular layout and is also relied on by the public extract_tables API for line-less PDFs. Set to false from callers that want the conservative “no ruling lines → no tables” behaviour (e.g. plain-text extraction paths that explicitly opt out — see extract_page_tables).

False-positive prose / TOC / underline tables that this default would previously have surfaced are filtered post-detection by the looks_like_prose_table shape gate and a ≥ 3-row evidence requirement on text-only and h-rule paths.

Default: true.

Implementations

impl TableDetectionConfig

pub fn strict() -> Self

Create a strict table detection configuration.

pub fn relaxed() -> Self

Create a relaxed table detection configuration.

Trait Implementations

impl Clone for TableDetectionConfig

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TableDetectionConfig

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

Formats the value using the given formatter.

impl Default for TableDetectionConfig

fn default() -> Self

Returns the “default value” for a type.

impl PartialEq for TableDetectionConfig

fn eq(&self, other: &Self) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for TableDetectionConfig