pter 0.1.0

use scraper::ElementRef;

/// Determine whether a `<table>` element is a data table or a layout table.
///
/// Email HTML overwhelmingly uses tables for layout. A table is considered
/// a **data table** if it has structural indicators of tabular data:
/// - Contains `<th>` elements
/// - Has a `<caption>` child
/// - Has `role="grid"` or `role="table"`
/// - Has multiple rows where multiple cells contain substantive text
///
/// Everything else is treated as a layout table and unwrapped.
pub fn is_data_table(table: ElementRef) -> bool {
    let el = table.value();

    // role attribute
    if let Some(role) = el.attr("role") {
        if role == "grid" || role == "table" {
            return true;
        }
        // role="presentation" is an explicit layout signal
        if role == "presentation" || role == "none" {
            return false;
        }
    }

    let mut has_th = false;
    let mut has_caption = false;
    let mut multi_cell_rows = 0u32;

    for descendant in table.descendants() {
        if let Some(el_ref) = ElementRef::wrap(descendant) {
            match el_ref.value().name() {
                "th" => has_th = true,
                "caption" => has_caption = true,
                "tr" => {
                    let cell_count = el_ref
                        .children()
                        .filter_map(ElementRef::wrap)
                        .filter(|c| {
                            let name = c.value().name();
                            (name == "td" || name == "th") && has_substantive_text(*c)
                        })
                        .count();
                    if cell_count > 1 {
                        multi_cell_rows += 1;
                    }
                }
                _ => {}
            }
        }
    }

    if has_th || has_caption {
        return true;
    }

    // Multiple rows with multiple substantive cells = data table
    multi_cell_rows >= 2
}

/// Check if an element contains meaningful text (not just whitespace/nbsp).
fn has_substantive_text(el: ElementRef) -> bool {
    let text = el.text().collect::<String>();
    let trimmed = text.trim().replace('\u{a0}', ""); // strip &nbsp;
    trimmed.len() > 1 // more than a single character
}

/// Extract rows and cells from a data table for markdown rendering.
///
/// Returns (headers, rows) where each is a Vec of cell text strings.
/// If no `<thead>`/`<th>` row exists, the first row is used as headers.
pub fn extract_table_data(table: ElementRef) -> (Vec<String>, Vec<Vec<String>>) {
    let mut headers: Vec<String> = Vec::new();
    let mut rows: Vec<Vec<String>> = Vec::new();

    // Look for thead/th first
    for descendant in table.children().filter_map(ElementRef::wrap) {
        let name = descendant.value().name();
        if name == "thead" {
            for tr in descendant.children().filter_map(ElementRef::wrap) {
                if tr.value().name() == "tr" {
                    headers = extract_cells(tr);
                    break; // first row of thead
                }
            }
        } else if name == "tbody" || name == "tr" {
            let trs: Box<dyn Iterator<Item = ElementRef>> = if name == "tbody" {
                Box::new(
                    descendant
                        .children()
                        .filter_map(ElementRef::wrap)
                        .filter(|e| e.value().name() == "tr"),
                )
            } else {
                Box::new(std::iter::once(descendant))
            };

            for tr in trs {
                let cells = extract_cells(tr);
                if !cells.is_empty() {
                    // If we haven't found headers yet and this row has <th> cells,
                    // treat it as the header row
                    if headers.is_empty() && has_th_cells(tr) {
                        headers = cells;
                    } else {
                        rows.push(cells);
                    }
                }
            }
        }
    }

    // If still no headers, promote first data row
    if headers.is_empty() && !rows.is_empty() {
        headers = rows.remove(0);
    }

    (headers, rows)
}

fn extract_cells(tr: ElementRef) -> Vec<String> {
    tr.children()
        .filter_map(ElementRef::wrap)
        .filter(|e| {
            let n = e.value().name();
            n == "td" || n == "th"
        })
        .map(|cell| {
            let text = cell.text().collect::<String>();
            text.split_whitespace().collect::<Vec<_>>().join(" ")
        })
        .collect()
}

fn has_th_cells(tr: ElementRef) -> bool {
    tr.children()
        .filter_map(ElementRef::wrap)
        .any(|e| e.value().name() == "th")
}

/// Render a data table as a GFM markdown table.
pub fn render_markdown_table(headers: &[String], rows: &[Vec<String>]) -> String {
    if headers.is_empty() {
        return String::new();
    }

    let col_count = headers.len();
    let mut out = String::new();

    // Header row
    out.push('|');
    for h in headers {
        out.push(' ');
        out.push_str(h);
        out.push_str(" |");
    }
    out.push('\n');

    // Separator row
    out.push('|');
    for _ in 0..col_count {
        out.push_str(" --- |");
    }
    out.push('\n');

    // Data rows
    for row in rows {
        out.push('|');
        for i in 0..col_count {
            out.push(' ');
            if let Some(cell) = row.get(i) {
                out.push_str(cell);
            }
            out.push_str(" |");
        }
        out.push('\n');
    }

    // Remove trailing newline (caller handles spacing)
    out.trim_end().to_string()
}

#[cfg(test)]
mod tests {
    use super::*;
    use scraper::{Html, Selector};

    fn parse_table(html: &str) -> Html {
        Html::parse_document(html)
    }

    fn select_table(doc: &Html) -> ElementRef<'_> {
        let sel = Selector::parse("table").unwrap();
        doc.select(&sel).next().unwrap()
    }

    #[test]
    fn single_cell_is_layout() {
        let doc = parse_table("<table><tr><td>content</td></tr></table>");
        assert!(!is_data_table(select_table(&doc)));
    }

    #[test]
    fn table_with_th_is_data() {
        let doc = parse_table(
            "<table><tr><th>Name</th><th>Age</th></tr><tr><td>Alice</td><td>30</td></tr></table>",
        );
        assert!(is_data_table(select_table(&doc)));
    }

    #[test]
    fn table_with_caption_is_data() {
        let doc = parse_table(
            "<table><caption>Users</caption><tr><td>Alice</td><td>30</td></tr></table>",
        );
        assert!(is_data_table(select_table(&doc)));
    }

    #[test]
    fn role_presentation_is_layout() {
        let doc = parse_table(
            r#"<table role="presentation"><tr><td>layout</td><td>stuff</td></tr></table>"#,
        );
        assert!(!is_data_table(select_table(&doc)));
    }

    #[test]
    fn role_grid_is_data() {
        let doc =
            parse_table(r#"<table role="grid"><tr><td>Alice</td><td>30</td></tr></table>"#);
        assert!(is_data_table(select_table(&doc)));
    }

    #[test]
    fn multi_row_multi_cell_is_data() {
        let doc = parse_table(
            "<table>\
            <tr><td>Alice</td><td>Engineer</td></tr>\
            <tr><td>Bob</td><td>Designer</td></tr>\
            </table>",
        );
        assert!(is_data_table(select_table(&doc)));
    }

    #[test]
    fn spacer_cells_not_substantive() {
        let doc = parse_table(
            "<table><tr><td>content</td><td>&nbsp;</td></tr>\
            <tr><td>more</td><td> </td></tr></table>",
        );
        // Only one substantive cell per row
        assert!(!is_data_table(select_table(&doc)));
    }

    #[test]
    fn render_simple_table() {
        let headers = vec!["Name".into(), "Age".into()];
        let rows = vec![
            vec!["Alice".into(), "30".into()],
            vec!["Bob".into(), "25".into()],
        ];
        let md = render_markdown_table(&headers, &rows);
        assert_eq!(
            md,
            "| Name | Age |\n| --- | --- |\n| Alice | 30 |\n| Bob | 25 |"
        );
    }

    #[test]
    fn render_empty_headers() {
        let md = render_markdown_table(&[], &[]);
        assert_eq!(md, "");
    }

    #[test]
    fn extract_with_thead() {
        let doc = parse_table(
            "<table><thead><tr><th>A</th><th>B</th></tr></thead>\
            <tbody><tr><td>1</td><td>2</td></tr></tbody></table>",
        );
        let (h, r) = extract_table_data(select_table(&doc));
        assert_eq!(h, vec!["A", "B"]);
        assert_eq!(r, vec![vec!["1".to_string(), "2".to_string()]]);
    }

    #[test]
    fn extract_promotes_first_row() {
        let doc = parse_table(
            "<table><tr><td>Name</td><td>Val</td></tr>\
            <tr><td>X</td><td>Y</td></tr></table>",
        );
        let (h, r) = extract_table_data(select_table(&doc));
        assert_eq!(h, vec!["Name", "Val"]);
        assert_eq!(r, vec![vec!["X".to_string(), "Y".to_string()]]);
    }

    // -- Boundary tests for is_data_table role handling --

    #[test]
    fn role_none_is_layout() {
        // role="none" → explicit layout signal. Catches L22 `||` mutation
        // (presentation OR none); without the ||, "none" wouldn't short-circuit.
        let doc = parse_table(
            r#"<table role="none"><tr><th>X</th><th>Y</th></tr><tr><td>1</td><td>2</td></tr></table>"#,
        );
        // Even with <th>, the explicit role="none" should win.
        assert!(!is_data_table(select_table(&doc)));
    }

    #[test]
    fn role_table_is_data() {
        // role="table" → data. Catches L22 == "grid" mutating to != (which would
        // make grid not match) AND covers the parallel `|| role == "table"` arm.
        let doc =
            parse_table(r#"<table role="table"><tr><td>a</td></tr></table>"#);
        assert!(is_data_table(select_table(&doc)));
    }

    #[test]
    fn role_unknown_falls_through_to_structural() {
        // Unknown role → no early decision; structural rules apply.
        // Single-cell single-row layout table → not data.
        let doc =
            parse_table(r#"<table role="banner"><tr><td>only one cell</td></tr></table>"#);
        assert!(!is_data_table(select_table(&doc)));
    }

    #[test]
    fn role_presentation_overrides_structure() {
        // role="presentation" → layout, even with multiple substantive rows.
        // Catches L22 == "presentation" mutating to != (which would skip this check).
        let doc = parse_table(
            r#"<table role="presentation"><tr><td>Alice</td><td>Engineer</td></tr>\
            <tr><td>Bob</td><td>Designer</td></tr></table>"#,
        );
        assert!(!is_data_table(select_table(&doc)));
    }

    // -- Boundary tests for has_substantive_text > 1 --

    #[test]
    fn single_char_cells_not_substantive() {
        // Two rows of single-char cells → not substantive → not a data table.
        // Catches L66 `>` mutating to `>=`: with >=, single chars become substantive
        // and these two rows would qualify as a data table.
        let doc = parse_table(
            "<table><tr><td>a</td><td>b</td></tr><tr><td>c</td><td>d</td></tr></table>",
        );
        assert!(!is_data_table(select_table(&doc)));
    }

    #[test]
    fn two_char_cells_are_substantive() {
        let doc = parse_table(
            "<table><tr><td>ab</td><td>cd</td></tr><tr><td>ef</td><td>gh</td></tr></table>",
        );
        assert!(is_data_table(select_table(&doc)));
    }

    // -- Boundary tests for extract_table_data tbody handling --

    #[test]
    fn extract_with_tbody_no_thead() {
        // Catches L87 `== "tbody"` mutating to != (which would skip tbody).
        let doc = parse_table(
            "<table><tbody><tr><td>Name</td><td>Val</td></tr><tr><td>X</td><td>Y</td></tr></tbody></table>",
        );
        let (h, r) = extract_table_data(select_table(&doc));
        // First tbody row promoted to headers; second row is data.
        assert_eq!(h, vec!["Name", "Val"]);
        assert_eq!(r, vec![vec!["X".to_string(), "Y".to_string()]]);
    }

    // -- Boundary tests for the headers-vs-th-row decision (L104 &&) --

    #[test]
    fn thead_present_blocks_later_th_row_promotion() {
        // Headers already set by thead. A later th-row should NOT overwrite them.
        // Catches L104 `&&` mutating to `||`: with ||, has_th_cells alone would
        // re-promote, clobbering the thead headers.
        let doc = parse_table(
            "<table><thead><tr><th>A</th><th>B</th></tr></thead>\
            <tbody><tr><th>X</th><th>Y</th></tr><tr><td>1</td><td>2</td></tr></tbody></table>",
        );
        let (h, r) = extract_table_data(select_table(&doc));
        assert_eq!(h, vec!["A", "B"], "thead headers must not be overwritten");
        // Both the th-row and the td-row become data rows.
        assert_eq!(r.len(), 2);
    }

    #[test]
    fn no_thead_th_row_promotes_to_headers() {
        // No thead, but a tr full of th cells → that tr's cells become headers.
        // Catches `has_th_cells -> bool` always-false mutation (which would
        // make this row become a data row instead).
        let doc = parse_table(
            "<table><tr><th>X</th><th>Y</th></tr><tr><td>1</td><td>2</td></tr></table>",
        );
        let (h, r) = extract_table_data(select_table(&doc));
        assert_eq!(h, vec!["X", "Y"]);
        assert_eq!(r, vec![vec!["1".to_string(), "2".to_string()]]);
    }

    #[test]
    fn all_td_rows_promote_first_to_headers() {
        // No th anywhere → has_th_cells is false for every row → first row promoted
        // by the `if headers.is_empty() && !rows.is_empty()` fallback.
        // Catches `has_th_cells -> bool` always-true mutation (which would promote
        // every row as headers, leaving rows empty after the first).
        let doc = parse_table(
            "<table><tr><td>Name</td><td>Val</td></tr><tr><td>X</td><td>Y</td></tr><tr><td>P</td><td>Q</td></tr></table>",
        );
        let (h, r) = extract_table_data(select_table(&doc));
        assert_eq!(h, vec!["Name", "Val"]);
        assert_eq!(r.len(), 2);
    }

    // -- Boundary test for has_th_cells (L139 == "th") --

    #[test]
    fn td_only_row_is_not_a_header_row() {
        // A tr with only <td> cells should NOT promote to headers when other
        // rows exist. Catches L139 `== "th"` mutating to `!=` (which would
        // match td cells and incorrectly treat every td row as a header row).
        let doc = parse_table(
            "<table><tr><td>data-1</td><td>data-2</td></tr>\
            <tr><td>data-3</td><td>data-4</td></tr>\
            <tr><td>data-5</td><td>data-6</td></tr></table>",
        );
        let (h, r) = extract_table_data(select_table(&doc));
        // First row is promoted (via the fallback at the end), leaving exactly two data rows.
        assert_eq!(h, vec!["data-1", "data-2"]);
        assert_eq!(r.len(), 2, "remaining rows should be data, not headers");
    }
}