tables are no longer recognized if a pipe is not put at the beginning
Closed this issue · 9 comments
Environment
Joplin version: 2.10.8
Platform: Android & Linux
Steps to reproduce
- create a note with a table, for instance:
column 1 | column 2
- | -
text 1 | text 2
Describe what you expected to happen
until version 2.9.17 this was recognized by the viewer as a table even if it does not have a pipe | at the beginning of every row.
Since version 2.10.8 a pipe must be present at least at the line that separate column's titles from rows, like this:
column 1 | column 2
| - | -
text 1 | text 2
As far as I know, the former was a "short way" to make a table in markdown, accepted also by other program (eg ReText), but I have no idea how much it's official
Are you sure it ever worked? Or maybe you had the table module enabled in the options? (Under the Markdown section)
100 % sure
I did not change options under the markdown section and "enable multimardown table extension" is set to false, but enabling it have the same behaviour as previous version. Probably, before the update to 2.10.8 that was enabled and disabled during/after the update
The GFM spec says:
A leading and trailing pipe is also recommended for clarity of reading, and if there’s otherwise parsing ambiguity.
So it should still be supported without leading and trailing pipes.
This is 2.8.8 without the table markdown extension:
This is pulled from dev:
You can see that codemirror actually realises it is a valid table but the HTML render doesn't which seems like a regression somewhere.
Broken due to this upstream change: markdown-it/markdown-it#767
It doesn't seem they are inclined to fix this upstream.
I confirmed that on Linux it does not work.
On macOS it still seems to work:
Joplin 2.10.10 (dev, darwin)
Client ID: e354a932e6e143c782264aad4112b674
Sync Version: 3
Profile Version: 42
Keychain Supported: Yes
Revision: d48a5efa0 (dev)
It doesn't seem they are inclined to fix this upstream.
I'm markdown-it maintainer, and I think I can answer what happened there.
Source of the issue
Github parses your code as a list, as demonstrated below:
a | b
- | -
1 | 2
a | b
- | -
1 | 2
This change in markdown-it is done to align markdown-it with cmark-gfm behavior. It seems to be much better if all parsers render your text the same way. But previously, if you copy-pasted your markdown into github, it'd just be broken.
Note that you don't have to have leading pipe, one extra minus would suffice:
column 1 | column 2
-- | --
text 1 | text 2
How to fix this
- I suggest to open an issue against https://github.com/github/cmark-gfm first, be sure to mention other parsers (ReText?) which behave differently. They are the maintainers of gfm spec, so it's their job to clarify it.
- If they change the spec or github parser, open a new issue in markdown-it to reflect those. Replying in an existing closed issue will easily get forgotten like it just happened with this case.
I believe the position of markdown-it team on this case is: we don't care how it is rendered. Just please make it behave the same way with all parsers.
You don't seem to consider that people's documents are broken now while the "fixes" you propose would take months to get anywhere, and most likely nothing will change since they aren't going to change the spec for us.
I feel the way this was handled was careless to be honest. The spec says this:
A leading and trailing pipe is also recommended for clarity of reading, and if there’s otherwise parsing ambiguity.
So it's only a recommendation, and just for this you potentially broke the documents of hundreds of people. I hope Markdown-it won't make too many such changes in the future just for a supposed correctness.
Closing as unfortunately there's not much not we can do about it