When there is only one "\n", the table header cannot be recognized.

[Yuanarcheannovice]

Steps to reproduce the problem (provide example Markdown if applicable):

" \n\n**aa**：  \n| ab   | bb     | cc     | dd     | ee     |\n|--------|----------|----------|----------|----------|\n| ab   | dff| dff| sf| sdfa|\n| sdf   | dfs| sdf| dsf| sdfs|\n| sdf   | sdf   | sdf       | ss   | d       |\n| ss   | dd   | ab       | bc   | 水       |\n"

Expected behavior:

Actual behavior:

It is in the Markdown style output by the AI.

(Also see what the reference implementation does: https://spec.commonmark.org/dingus/)

[robinst]

A more minimal example of what this is about (I believe):

text
|a  |
|---|
|b  |

GitHub renders it as a paragraph and a table, commonmark-java just as a paragraph.

The difficult part here is that the table needs to interrupt the paragraph, but we only know it's a table by the third line (|---|).. Note that GitHub's spec about tables doesn't mention this case, but anyway. We should look into whether this can be supported. It might need changes in the extension APIs.

[az4mxl]

public static class Factory extends AbstractBlockParserFactory {

    @Override
    public BlockStart tryStart(ParserState state, MatchedBlockParser matchedBlockParser) {
        CharSequence line = state.getLine();
        CharSequence paragraph = matchedBlockParser.getParagraphContent();
        if (paragraph != null && paragraph.toString().contains("|") && !paragraph.toString().contains("\n")) {
            CharSequence separatorLine = line.subSequence(state.getIndex(), line.length());
            List<TableCell.Alignment> columns = parseSeparator(separatorLine);
            if (columns != null && !columns.isEmpty()) {
                List<String> headerCells = split(paragraph);
                if (columns.size() >= headerCells.size()) {
                    return BlockStart.of(new TableBlockParser(columns, headerCells))
                            .atIndex(state.getIndex())
                            .replaceActiveBlockParser();
                }
            }
        }

        try {
            // 兼容下面这个格式，列表中有表格，表格前面只有一个换行符
            // * 列表第一行\n* 列表第二行\n| 名称 | 生卒年份 | 主要作品 |\n|------|----------|------|----------|\n| 李白 | 701-762 | 《静夜思》《将进酒》 |
            if (matchedBlockParser.getMatchedBlockParser() instanceof ListBlockParser) {
                CharSequence pbOrigin = ((ParagraphParser) (state.getActiveBlockParser())).getContentString();// 上一个ListItem中的数据
                if (pbOrigin != null && pbOrigin.toString().contains("\n")) {
                    String[] pbs = pbOrigin.toString().split("\n");
                    if (pbs.length == 2) {
                        String pb = pbs[pbs.length - 1];// 拿到表格头
                        CharSequence separatorLine = line.subSequence(state.getIndex(), line.length());
                        List<TableCell.Alignment> columns = parseSeparator(separatorLine);
                        if (columns != null && !columns.isEmpty()) {
                            List<String> headerCells = split(pb);
                            if (columns.size() >= headerCells.size()) {
                                StringBuilder pbSB = (StringBuilder) pbOrigin;
                                deleteStringEndText(pbSB, pb);// 删除上一个ListItem中的表格头
                                return BlockStart.of(new TableBlockParser(columns, headerCells))
                                        .atIndex(state.getIndex());
                            }
                        }
                    }
                }
            }
        } catch (Exception e) {
            e.printStackTrace();
        }

        return BlockStart.none();
    }
}

private static void deleteStringEndText(StringBuilder sb, CharSequence suffix){
    int len = suffix.length();
    if (sb.length() >= len && sb.substring(sb.length() - len).equals(suffix)) {
        sb.delete(sb.length() - len, sb.length());
    }
}

[robinst]

Released in 0.25.0 now: https://github.com/commonmark/commonmark-java/releases/tag/commonmark-parent-0.25.0