[Bug]: Markdown tables missing outer pipe delimiters - causes parser compatibility issues

[PatD42]

crawl4ai version

0.8.0 (also confirmed in 0.7.4)

Expected Behavior

Markdown tables should include leading and trailing pipe (|) delimiters on every row for maximum compatibility with markdown parsers. While this is technically valid GitHub Flavored Markdown (GFM), it causes parsing issues with many markdown processors and violates the recommended best practice for compatibility.

Current Behavior

Parameter	Guideline	Common Sources	Health Considerations	Comments
Arsenic (2006)	0.010 ALARA	Naturally occurring	Cancer risk	Keep as low as possible

Issues this causes:

1. Many markdown parsers fail to render these tables correctly (Incorrect parsing of GFM tables with missing leading/trailing pipes showdownjs/showdown#230)
2. Static site generators (Jekyll, Hugo) have inconsistent rendering
3. Some IDEs and markdown previewers don't recognize the table structure
4. LLM processing may have reduced accuracy with inconsistent formatting
5. Potential data loss when parsers misalign columns

Is this reproducible?

Yes

Inputs Causing the Bug

Any webpage containing HTML `<table>` elements will produce markdown tables with missing outer pipes.

**Test URL used:**
https://www.canada.ca/en/health-canada/services/environmental-workplace-health/reports-publications/water-quality/guidelines-canadian-drinking-water-quality-summary-table.html

This page contains multiple tables with 5-7 columns and 100+ rows. All tables exhibit the issue.

Steps to Reproduce

1. Install crawl4ai 0.8.0
  2. Crawl any webpage with HTML tables
  3. Examine the generated markdown
  4. Observe that table rows do not start/end with `|`

  **Specific test:**
  - Crawl the Canadian drinking water quality page
  - Check Table 2 (Chemical Parameters) - 97 rows, all missing outer pipes
  - Verify separator row format: `---|---|---|---|---` instead of `|---|---|---|---|---|`

Code snippets

import asyncio
  from crawl4ai import AsyncWebCrawler, CrawlerRunConfig, CacheMode

  async def test_table_format():
      """Test script demonstrating missing outer pipes in markdown tables"""

      url = "https://www.canada.ca/en/health-canada/services/environmental-workplace-health/reports-publications/water-quality/guidelines-canadian-drinking-water-quality-summary-table.html"

      run_config = CrawlerRunConfig(
          cache_mode=CacheMode.BYPASS,
          wait_until="networkidle",
          delay_before_return_html=2.0,
          excluded_tags=['nav', 'header', 'footer', 'form'],
          word_count_threshold=10
      )

      async with AsyncWebCrawler() as crawler:
          result = await crawler.arun(url=url, config=run_config)

          # Analyze table format
          lines = result.markdown.split('\n')
          table_rows_found = 0
          rows_missing_pipes = 0

          for i, line in enumerate(lines):
              if '|' in line and line.strip():
                  table_rows_found += 1
                  starts_with_pipe = line.strip().startswith('|')
                  ends_with_pipe = line.strip().endswith('|')

                  if not starts_with_pipe or not ends_with_pipe:
                      rows_missing_pipes += 1

                      # Print first few examples
                      if rows_missing_pipes <= 3:
                          print(f"\nLine {i}: {repr(line[:100])}")
                          print(f"  Starts with |: {starts_with_pipe}")
                          print(f"  Ends with |: {ends_with_pipe}")

          print(f"\n\nSummary:")
          print(f"  Total table rows: {table_rows_found}")
          print(f"  Rows missing outer pipes: {rows_missing_pipes}")
          print(f"  Percentage affected: {rows_missing_pipes/table_rows_found*100:.1f}%")

  asyncio.run(test_table_format())

OS

Mac OS

Python version

3.13.11

Browser

n/a

Browser version

n/a

Error logs & Screenshots (if applicable)

No response

[Ahmed-Tawfik94]

Root Cause

The table formatting logic in CustomHTML2Text (which inherits from HTML2Text) does not implement the GFM standard for table boundaries:

1. Missing Leading Pipe: First cell in each row doesn't output a leading |
2. Missing Trailing Pipe: Rows don't output a trailing | after the last cell
3. Broken Separator Format: Separator rows join dashes with pipes but lack boundary pipes

Flow Diagram

HTML Table
  └─ <tr> START
      └─ td_count = 0
      └─ split_next_td = False
      └─ <td> START
          └─ split_next_td is False → NO "| " output
          └─ split_next_td = True
          └─ [CELL CONTENT OUTPUT]
      └─ <td> END
      └─ <td> START
          └─ split_next_td is True → "| " output
          └─ split_next_td = True
          └─ [CELL CONTENT OUTPUT]
      └─ <td> END
  └─ <tr> END
      └─ split_next_td = False
      └─ soft_br() called
      └─ NO TRAILING "|" OUTPUT

Result: Content1 | Content2 instead of | Content1 | Content2 |

---

Solution

Fix: Add Leading and Trailing Pipes to Table Rows

Modify /Users/ahmet/work/crawl4ai/crawl4ai/html2text/__init__.py in the table handling section (around line 718):

Original Code (Broken):

if tag in ["td", "th"] and start:
    if self.split_next_td:
        self.o("| ")
    self.split_next_td = True

Fixed Code:

if tag in ["td", "th"] and start:
    # Always output a leading pipe for first cell and between cells
    self.o("| ")
    self.split_next_td = True

And fix the separator row:

Original:

if tag == "tr" and not start and self.table_start:
    # Underline table header
    self.o("|".join(["---"] * self.td_count))
    self.soft_br()
    self.table_start = False

Fixed:

if tag == "tr" and not start and self.table_start:
    # Underline table header with proper GFM format
    self.o("| " + " | ".join(["---"] * self.td_count) + " |")
    self.soft_br()
    self.table_start = False

And add trailing pipe at row end:

if tag == "tr" and not start:
    self.o(" |")  # Add trailing pipe before line break
    self.split_next_td = False
    self.soft_br()

---

[vikas-gits-good]

Can i take this issue?
And can I reuse the code you suggested as a fix?