I don't see table headers used a lot in wikipedia, so maybe its own markup could become too confusing. Could that be solved by

{| |class=header|The header text |- |Just a cell |}

Is there a special function for <th> anyway, or is it just the visual difference? In that case, we wouldn't need <th> at all, but could do with a <td> CSS class (like above).

But, if there's a demand for a <th> equivalent, I'll add one. Someone suggested using == Headers == in the table markup makes <th> tags.

Summary of possible solutions: 1. Ignore <th> completely 2. Make a special tag ("!" or "|:" or something) 3. Use "| == Text ==" as "<th>Text</th>"

What's it gonna be? :-)

Magnus

P.S.: Seems someone wiped the stylesheet definitions for the floating-right image from the test site. Too bad.

Gutza wrote:

How about adding support for

{| ! Table header 1 ! Table header 2 |- | Table content 1 | Table content 2 |}

To generate proper table headers as specified by HTML 4.0? Also, you could start a header row using

{| !- | Table header 1 | Table header 2 |- | Table content 1 | Table content 2 |}

because a proper implementation of the markup shouldn't echo the first

<TR> only after encountering the first content tag, so encountering a "!-" before that could echo the proper HTML header tags.

Gutza

Magnus Manske wrote:

...

OK, anyone opposed to

• wiki table markup *in general*
• the *markup style* I use at test.wikipedia.org
• my *implementation* of that style

please speak up now! Otherwise, I'll commit it to the stable branch of the software, which means it will go online with the next update.

Magnus

---

Wikipedia-l mailing list Wikipedia-l@Wikipedia.org http://mail.wikipedia.org/mailman/listinfo/wikipedia-l

---

Wikipedia-l mailing list Wikipedia-l@Wikipedia.org http://mail.wikipedia.org/mailman/listinfo/wikipedia-l

Perhaps give what you got a trial run, and let people complain or not as they come. First add a front page news blurb (like MIT) linking to the Documentation. Allow for comment there. (faster feedback loop)

PS- will your magical program perhaps attempt someday to convert html table markup to Wiki? I assume this is unlikely.

~S~

--- Peter Gervai grin@tolna.net wrote:

On Mon, Sep 15, 2003 at 04:47:05PM +0200, Magnus Manske wrote:

...

Is there a special function for <th> anyway, or is

it just the visual

...

difference?

It _may_ get rendered differently by non-graphical renderers, like those for the blind. I believe it's nice behaviour to use them for headers. _______________________________________________ Wikipedia-l mailing list Wikipedia-l@Wikipedia.org

http://mail.wikipedia.org/mailman/listinfo/wikipedia-l

__________________________________ Do you Yahoo!? Yahoo! SiteBuilder - Free, easy-to-use web site design software http://sitebuilder.yahoo.com

Magnus Manske wrote:

Is there a special function for <th> anyway, or is it just the visual difference?

Please note there are two ways of specifying the table headers: TH cells (in which case you must specify a scope attribute) and the THEAD row group. The THEAD and TBODY groups are mandatory in HTML 4.x strict IIRC -- but that's not the reason why I think they should be supported in itself. I think they would be interesting to support for the same reasons why they have been added to HTML:

1. Separation of content and format by allowing designers to specify the content in predefined tags; 2. Accessibility support and special applications support; 3. Greater user agent customization based on the structure information (e.g. future browsers will probably render the headers of a table on each page the table spans on).

Please note that, should you implement this feature, you'll need to find a way to properly retrieve and render the headers in HTML:

<TABLE> <TR> <TH>Header 1</TH> <TH>Header 2</TH> </TR> <TR> <TD>Content 1</TD> <TD>Content 2</TD> </TR> </TABLE>

is incorrect, it should be either

<TABLE> <TR> <TH SCOPE='column'>Header 1</TH> <TH SCOPE='column'>Header 2</TH> </TR> <TR> <TD>Content 1</TD> <TD>Content 2</TD> </TR> </TABLE>

or ideally

<TABLE> <THEAD> <TR> <TD>Header 1</TH> <TD>Header 2</TH> </TR> </THEAD> <TBODY> <TR> <TD>Content 1</TD> <TD>Content 2</TD> </TR> </TBODY> </TABLE>

if possible. Also, <TH SCOPE='row'> should be used where it can be determined that the contributor wanted to render a vertical header:

<TABLE> <TBODY> <TR> <TH SCOPE='row'>Header 1</TH> <TD>Content 1</TD> </TR> <TR> <TH SCOPE='row'>Header 2</TH> <TD>Content 2</TD> </TR> </TBODY> </TABLE>

Rather than using a CSS class for headers, I'd prefer not using them at all, because it would generate the same type of structure/content/rendering mix the authors of the HTML specification wanted to avoid.

I don't want to impose my notation in your implementation of the markup, but I personally find the exclamation mark to be a better solution than "|:" exactly because I personally find it to be more intuitive and less confusing: it looks similar to the vertical bar, it uses the same one-character convention for opening cells as the vertical bar uses and the exclamation mark is generally associated with important data, which a header in a table is.

Absolutely obviously indubitably just my 2c. :)

Gutza

On Mon, 15 Sep 2003 19:35:47 +0200, Schneelocke wikipedia-l@gl00on.net gave utterance to the following:

...

...

Is there a special function for <th> anyway, or is it just the visual difference?

Please note there are two ways of specifying the table headers: TH cells (in which case you must specify a scope attribute) and the THEAD row group. The THEAD and TBODY groups are mandatory in HTML 4.x strict IIRC [...]

In XHTML 1.1 at least, thead, tfoot and tbody are optional [1]. In HTML 4.01, it seems to me that tbody at least is required, while thead and tfoot are optional [2].

http://www.w3.org/TR/xhtml-modularization/abstract_modules.html#s_tablemodul... 2. http://www.w3.org/TR/html4/sgml/dtd.html

There is no validation error from absence of TBODY, which is fortunate given that 99.999% of tables on HTML 4 pages do not include it. I think its status is IMPLIED.

TFOOT is a dangerous element to use. It must appear in the source before TBODY, but in non-supporting browsers (of which there are a number in common use) that means that your footer rows will appear at the top of your table rather than the bottom (one of the few instances where a new HTML element does not degrade gracefully).

As for the recomemndation to use the scope and headers attributes, there isn't one browser which supports them yet. They are a nice theory, but no practical use.

I would opt for a means of generating TH - which must be available on any cell (I have tables with headers on the left, or a smaller subsection in one corner.

Ulrich Fuchs wrote:

Am Montag, 15. September 2003 15:51 schrieb Magnus Manske:

...

OK, anyone opposed to

• wiki table markup *in general*
• the *markup style* I use at test.wikipedia.org
• my *implementation* of that style

please speak up now! Otherwise, I'll commit it to the stable branch of the software, which means it will go online with the next update.

Ok, just speaking up a little bit, but just repeating what has been said before.

I doubt that a new markup for tables will make things a lot more easier for newbies. Tables *are* a complex thing, and they are difficult to read in any markup language, because cells rendered *beside* the other are coded *below* each other in the source (at least if there is som more content requireing line breaks). So I doubt that a newbie will be able to edit a table like it's used in the country or element articles, no matter what markup we are using.

Therefore I really liked the idea of having a "table"-namespace, and a way of importing the tables in that namespace as easiliy as we do with images right now. The article source would be cleared of the table, and the table might be as complex as needed without scaring away the newbies.

However, I can imagine that the new markup is a fine way for rendering *simple* tables within the article, giving newbies a chance to do simple table layouts the easy way (there are a lot of issues that can be aligned in simple tables, so the feature *is* useful). It's somewhat more readable, easier to understand and therefore easier to write than html.

But, as I said, for large and complex tables that new markup will not solve our problems.

Uli

Agreed, for large tables it probably wouldn't as it stands. But if your worries are the newbies, a table creation tool could be added in Wikipedia, which would make things easier with newbies. I do however think the new table markup could be useful for newbies because it wouldn't require strict stack-like (FIFO) open/close tags which I guess are the main drawback of HTML tables; also, with the advent of proper table header wiki tags, newbies could easily properly format relatively simple tables, without having to go through freakish "<TD BGCOLOR='#DDDDDD'>" tags. And finally, being a newbie doesn't equate being an idiot, people can learn even if the cells rendered beside one another must be coded under one another, that's relatively easy to understand, given that people generally read left-to-right/top-to-bottom, so it's no change of order, although admittedly there is a considerable change in format (I expect there will be some support for right-to-left languages as well).

Gutza

Tomasz Wegrzanowski wrote:

On Mon, Sep 15, 2003 at 03:51:46PM +0200, Magnus Manske wrote:

...

OK, anyone opposed to

• wiki table markup *in general*
• the *markup style* I use at test.wikipedia.org
• my *implementation* of that style

please speak up now! Otherwise, I'll commit it to the stable branch of the software, which means it will go online with the next update.

I'm not a great fan of it. Can you provide complete specification of how is it supposed to behave ? It seems to be stealing whole 3 characters - {, }, and |, and to be much more ambiguous than it should.

It only steals one character--"{"--and at that, only at the beginning of a paragraph. Not bad if you ask me (not that you did), 'cause I don't think there are so many paragraphs in Wikipedia starting with a curly bracket. It does "steal" the other characters you mention, plus the exclamation mark I suggested, at the beginning of a paragraph *within* a table. But one you open a table, you're aware those are mandatory, so no worries there.

Gutza

Tomasz Wegrzanowski wrote:

On Mon, Sep 15, 2003 at 11:47:27PM +0300, Gutza wrote:

...

Tomasz Wegrzanowski wrote:

...

On Mon, Sep 15, 2003 at 03:51:46PM +0200, Magnus Manske wrote: I'm not a great fan of it. Can you provide complete specification of how is it supposed to behave ? It seems to be stealing whole 3 characters - {, }, and |, and to be much more ambiguous than it should.

It only steals one character--"{"--and at that, only at the beginning of a paragraph. Not bad if you ask me (not that you did), 'cause I don't think there are so many paragraphs in Wikipedia starting with a curly bracket. It does "steal" the other characters you mention, plus the exclamation mark I suggested, at the beginning of a paragraph *within* a table. But one you open a table, you're aware those are mandatory, so no worries there.

Table is not like the math mode, which is completely separate from normal wiki syntax, but something inside what all the wiki syntax is supposed to be usable. And for something like that, it steals a bit too many characters.

How should I put it? It's *ONE* character *COMBINATION*! If one character combination is too many, well, I don't know, maybe you're right. I don't think it is too many though. Do you understand what I mean about it only stealing one character combination? If you don't start a table by using "{|" at the beginning of a line, "{" or "|" at the beginning of a line would be rendered as-is. Using "{", "|" or even "{|" anywhere else but the beginning of a line would be rendered as-is as well. Maybe I'm missing something though, please clarify your position if I overlooked something.

And what about precise description of this markup ?

Given the fact that, if you read the rest of this thread, this implementation is currently under discussion, I wouldn't expect the developer to start typing a description just yet.

Gutza

On Mon, 2003-09-15 at 18:49, Tomasz Wegrzanowski wrote:

.... We already had a bug where ;/: and : for namespaces conflicted (and btw., is it fixed ?).

Which bug is that?

-- brion vibber (brion @ pobox.com)

On Mon, 2003-09-15 at 19:43, Tomasz Wegrzanowski wrote:

On Mon, Sep 15, 2003 at 07:32:58PM -0700, Brion Vibber wrote:

...

On Mon, 2003-09-15 at 18:49, Tomasz Wegrzanowski wrote:

...

.... We already had a bug where ;/: and : for namespaces conflicted (and btw., is it fixed ?).

Which bug is that?

This one - http://test.wikipedia.org/wiki/A_bug (still there)

Blecch! Is there any reason not to remove the single-line ;: directory list syntax?

For the uninitiated: a line like this:

; foo : bar

would produce code something like:

<dl> <dt>foo</dt> <dd>bar</dd> </dl>

(but uglier, because our parser produces ugly code) which renders something like:

foo bar

Currently this triggers on :s in URLs and wikilinks as well as plain text, which is unexpected.

Requiring the : portion to be a second line would prevent this from being an issue.

-- brion vibber (brion @ pobox.com)

Brion Vibber schrieb:

Blecch! Is there any reason not to remove the single-line ;: directory list syntax?

For the uninitiated: a line like this:

; foo : bar Requiring the : portion to be a second line would prevent this from being an issue.

Well, you can write ; foo : bar currently, and it is the same result as you get with ; foo : bar - I like this parser :)

I think it can be changed, if we support the conversion.

Smurf

Magnus Manske wrote:

Tomasz Wegrzanowski wrote:

...

It's not only that - `|' becomes magical not only at the begining of line (what would be relatively harmless), but also inside it.

It will be only magical if

• it is used inside a table and
• if a '|' is the first char at the beginning of the line

Or a "!", as of latest (yes, <th> has finally arrived!)

The only use of '|' I can think of in text (not counting links, they won't be affected) is for ASCII drawings, and for these, the line always starts with a blank anyway.

...

Magnus said he's going to put it live on all pedias. That's definitely the time for giving exact description.

I put one up at http://test.wikipedia.org/wiki/Wiki_table_markup

Beautiful! I think this will make the life of newbies a lot easier when having to use tables. And I'm also happy with the <th> tags, it's way better than using classes; we just need to implement proper visual support for headers in CSS for everyone to use it (<th> should have a different background than <td> in all skins except printer-friendly IMHO).

Gutza

Gutza wrote:

Magnus Manske wrote: On a side note, this should probably also be supported in a fully-flexible implementation:

{| ! Header1 || Cell1 |- ! Header2 || Cell2 |}

IMHO most use of headers occurs in the very first line, so a line with multiple cells that starts with "!" will only have headers.

Your above example would have to be done like {| ! Header1 | Cell1 |- ! Header2 | Cell2 |}

Also: Currently, I do "||" parsing via "explode" (with prior replacing "!!" with "||" if line starts with "!"). Parsing the different tags would thus result in more code, and if you ever read my code, you don't want more of it than absolutely necessary ;-)

Magnus