#bug

@Mario Vavti Disabling the MarkdownPoster addon fixes this. Looks like some conflict with the addon.

@Mario Vavti @Saiwal

It seems to be a problem of URL in correlation with line feed.

If you place a URL in the code block (code tags or with Markdown between ``) and there is no line feed afterwards, i.e. no new line within the code block, then the URL is displayed correctly. If there is a line feed, the bbCode URL tags are displayed.

It seems to be somehow due to the parsing order. A line feed seems to cause the URL tags to be inserted. The parsing of the code tags then takes place afterwards and ensures that the URL tags are displayed as code.

@Der Pepe (Hubzilla) ⁂ ⚝ You're on to something here. I think the best we can do for now (apart from rewriting the entire bbcode parser) is to add all the test cases we can come up with, and try to make them all pass.

(... but that rewrite seems very tempting, tbh :)

I heard that Streams changed theirs so that it supports multiple markup languages, such as BBCode, Markdown, and perhaps HTML without additional addons. I wonder if that is worth porting to Hubzilla.

@Scott M. Stolz Was actually thinking about the same thing. I can have a look.

In any case, it would be nice if we could store the source text unmodified as typed. Now it will be parsed to an "optimized" bbcode before being submitted. So when editing a post it may not look the same as when you typed it in.

@Harald Eilertsen @Mario Vavti @Saiwal

I don't think a rewrite is required for THIS error (URL in code block).

Since no further markdown or bbCode parsing should (and may) take place within a code block, but only a predefined fixed font is used, it would be sufficient to search for the occurrence of the opening code tag first, as well as for the closing tag. And the content in between is ignored by the parser (except for the fixed font formatting).

@Harald Eilertsen True. And that is probably why editing breaks certain things, like ordered and unordered lists.

@Der Pepe (Hubzilla) ⁂ ⚝ Yeah, that's how you normally would write a parser, but the bbcode parser does not work like that. It uses regex to look for various tags and their contents, and then replace them with the html it should represent. For code blocks in particular, it handles them in two stages, first by replacing them with a base64 representation of their content (to prevent their content to be parsed further), and then at the end the content is then base64 decoded again.

It's inefficient, but works most of the time.

The real problems occur when you mix the Markdown parser into this mix. It does its thing before the bbcode parser, and works by parsing the markdown to html, and then the html back to bbcode. Which is then passed through the bbcode parser again.

Again, inefficient, but it works well enough most of the time.

At least as long as the input is well formed.

What I think is happening in @Saiwal's example, is that Markdown will recognize the URL, and create a HTML link for it. This will then be converted to a bbcode link again. The fact that it is within a code block is not relevant yet, because that will only be handled later when the content is passed to the bbcode parser again. At that time, the bbcode link will just be treated as verbatim text, instead of actual bbcode.

I'm still trying to reproduce this in a test case, so some details may be inaccurate.

@Harald Eilertsen Yes, that could really be the reason. The best way to solve this would probably be to throw the Markdown parser as an add-on overboard, as well as the bbCode parser, and really create a parser for bbCode and Markdown at the same time from scratch.

I wonder if it would be easier to use plain text as a default. Then have an option to select the parser you want. Whether it's bbcode, markdown, html, or otherwise. That way you don't have one piece of content running through multiple parsers. That would require people to select their parser on a per post basis, or have general global setting.

Just an idea. I have zero skills to implement such a thing.

@Grouchy Sometimes it is useful to mix markup. For example, there are some things that can only be done in HTML or BBCode, and it is a lot more convenient to use Markdown for things like bullets and ordered lists.

For the Hubzilla Wiki, I often use a mix of HTML and Markdown for that very reason. In posts, I usually mix BBCode and Markdown.

I am not a big fan of mixed content because it will not be parseable by anything else...

@Scott M. Stolz

That's understandable. For me, I want to write in just one, and not have another parser misinterpreting what I wrote.

@Mario Vavti In the end, it is HTML that is delivered to the end user. The main reason we use BBCode and Markdown is to make it easier for people posting, and to prevent people from using destructive HTML tags in their post.

So no one has to parse it except ourselves, because we are basically outputting filtered HTML to everyone else.

A lot of this parsing discussion revolves around the fact that the default interface does not use a rich text editor (WYSIWYG interface).

@Mario Vavti It certainly complicates things...

@Scott M. Stolz iirc friendica parses bbcode where possible. We deliver both BBcode and HTML. BBcode is especially useful for various observer specific codes. It is rendered at the receivers side and therefore can be adapted as needed.