[Bug] Markdown Table With one Column missing alignment fails to convert

[iJungleboy]

Describe the bug

In the latest release (just a few hours ago) something broke converting tables.
In the following example, the first table converts as expected the second does not

This works:

| Field       | PersonShared  | Person      |
| ---         | ---           | ---         |
| Name        | 🔊 Master     | 🔗 Inherit  |

While this fails - note the missing "---" under the second column "PersonShared"

| Field       | PersonShared  | Person      |
| ---         |               | ---         |
| Name        | 🔊 Master     | 🔗 Inherit  |

The first table converts as expected, the second results as the markdown being added to the page:

<p>While this fails - note the missing &quot;---&quot; under the second column &quot;PersonShared&quot;</p>
<p>| Field       | PersonShared  | Person      |
| ---         |               | ---         |
| Name        | 🔊 Master     | 🔗 Inherit  |</p>

Expected behavior

Both should convert - this did work before.

Context (please complete the following information):

• OS: Windows
• Docfx version: 2.78.6

Additional context

I'm quite certain this is a regression from #11088

[vicancy]

Thanks for reporting this. We traced the change to the Markdig update from 1.0.0 to 1.1.0 in Docfx 2.78.6. The relevant upstream change is Markdig's separator-validation fix.

The updated parser follows the GFM table specification: each separator cell needs hyphens (-), while the colons used to specify alignment are optional. Earlier versions accepted empty separator cells.

We recommend updating the affected tables by adding --- to those empty cells:

| Field | PersonShared | Person |
| --- | --- | --- |
| Name | 🔊 Master | 🔗 Inherit |

We appreciate that this changes behavior for existing documents. Would updating the affected tables be practical for your documentation? If you have concerns or constraints that would make this difficult, please share them so we can discuss further.

[iJungleboy]

@vicancy I already fixed my docs, so it doesn't matter.

But I'm pretty sure that others may not expect this, and I'm almost certain that this is a mistake.

I understand that a "Pipe-only row" as is referenced in the xoofx/markdig@7959e3b should be skipped, but omitting dashes in a field is almost certainly not the case they were trying to avoid.

So I'm good, but I assume others won't even notice that things break.

[vicancy]

Good point. I've opened xoofx/markdig#955 to allow empty separator cells again while keeping the pipe-only fix. Let's see what the Markdig maintainers think.