Chain style html markup - html

Motivation and Problem
There are several libraries for generating html markup strings using ruby (erb, haml, builder, markaby, tagz, ...), but I am not satisfied with any of them. The reason is that, except for erb, they take the nesting style rather than the chain style. And erb is a way to embed ruby within html rather than generating html with ruby.
To my understanding, one beauty of ruby lies in encouraging the use of chain style:
receiver.method1(args1).method2(args2). ... method_n(args_n)
instead of doing a nesting style:
method_n(...method2(method1(receiver, args1), args2), ... args_n)
But libraries mentioned above (except for erb) take the nesting style (sometimes with the help of block arguments).
My Idea
For my own purpose, I wrote a method dom so that I can do html markup in a chain style. When applied to a string, this example
"This is a link to SO".dom(:a, href: "http://stackoverflow.com").dom(:body).dom(:html)
will generate:
<html><body><a href="http://stackoverflow.com";>This is a link to SO</a></body></html>
When applied to an array, this:
[
["a".dom(:td), "b".dom(:td)].dom(:tr),
["c".dom(:td), "d".dom(:td)].dom(:tr)
].dom(:table, border: 1)
will generate
<table border="1";>
<tr>
<td>"a"</td>
<td>"b"</td>
</tr>
<tr>
<td>"c"</td>
<td>"d"</td>
</tr>
<table>
And, when applied without an explicit receiver (outside of a domain of strings and arrays),
dom(:img, src: "picture.jpg", width: 48, height: 48)
will generate
<img src="picture.jpg";width="48";height="48";/>
Note that all is done with just one method dom. This is much simpler than using other libraries. It is also flexible in that it is not affected by a change in the inventory of html tags; you just specify that with a symbol argument. In other libraries, there are classes and/or methods for each tag. Furthuremore, unlike erb, it is pure ruby. It is not a DSL that needs to be convertred.
My Implementation
The implementation is as follows:
class Hash
def attribute
map{|k, v| %Q{#{k}#{
case v
when TrueClass; ''
when Hash; %Q{="#{v.subattribute}"}
else %Q{="#{v}"}
end
;}}}.join
end
def subattribute
map{|k, v| %Q{#{k}:#{v};}}.join
end
end
class Array
def dom type, hash = {}
"<#{type} #{hash.attribute}>\n#{join("\n").gsub(/^/, " ")}\n</#{type}>"
end
end
class String
def dom type, hash = {}
"<#{type} #{hash.attribute}>#{self}</#{type}>"
end
end
class Object
def dom type, hash = {}
"<#{type} #{hash.attribute}/>"
end
end
Questions
Are there already stable libraries that do a similar thing?
What will be the potential problems to this approach (particularly to my implementation or to doing this in chain approach)?
Some attributes take boolean values, which are often encouraged to be omitted. For example, <input type="text";readonly> instead of <input type="text";readonly="true">. In my present implementation, I can do that by passing true (which will not be used in the end) as the value for such attribute like dom(:input, type: "text", readonly: true), but that seems redundant and is also part of the reason that I have case statement in the code, making it slower. Is there a better way to do this?
Are there any possible improvements to the implementation?

The main reason of haml and most other nested stuff is that it is basically easy to look how your HTML is nested. I am assuming this but do you actually code HTML or do you do more backend stuff? The reason for it is that in nested style, you will see how your elements are nested(which is important to you if you are also writing the styles)
While it is quite easier to write
.dom(:body).dom(:html)
it is difficult for a designer to see how the HTML flows without mapping and trying it to visualize in his head, whereas:
%html
%body
does that already with just one look.
Take not of a longer example:
"This is a link to SO".dom(:a, href: "http://stackoverflow.com").dom(:body).dom(:html)
Would it be easier for you if say the client or you needed to add a clickable image to the link? How would you have done it? In haml it is easy as:
%html
%body
%a{:href => "blah"}
= image_tag("image.png")
Also, IMHO, writing dom(:) for each html tag is just as tedious as writing the closing tag for it(which HAML and the others fixed)
Again, these are just my opinions from an XHTML/CSS programmer(as opposed to a ruby programmer perspective)
And lastly, I would also think that this can be turned into a community wiki or something as this doesn't merit an exact answer and will probably spawn a lot of subjective ones like this one.

Related

How to Create an HTML Template?

Problem
I have a collection of images with linked captions on a page. I want them each to have identical HTML.
Typically, i copy and paste the HTML over and over for each item. The problem is, if i want to tweak the HTML, i have to do it for all of them. It's time-consuming, and there's risk of mistakes.
Quick and Dirty Templating
I'd like to write just one copy of the HTML, list the content items as plain text, and on page-render the HTML would get automatically repeated for each content-item.
HTML
<p><img src=IMAGE-URL>
<br>
<a target='_blank' href=LINK-URL>CAPTION</a></p>
Content List
IMAGE-URL, LINK-URL, CAPTION
/data/khang.jpg, https://khangssite.com, Khang Le
/data/sam.jpg, https://samssite.com, Sam Smith
/data/joy.jpg, https://joyssite.com, Joy Jones
/data/sue.jpg, https://suessite.com, Sue Sneed
/data/dog.jpg, https://dogssite.com, Brown Dog
/data/cat.jpg, https://catssite.com, Black Cat
Single Item
Ideally, i could put the plain-text content for a single item anywhere on a page, with some kind of identifier to indicate which HTML template to use (similar to classes with CSS).
TEMPLATE=MyTemplate1, IMAGE-URL=khang.jpg, LINK-URL=https://khangssite.com, CAPTION=Khang Le
Implementation
Templating systems are widely used, like Django and Smarty on the server side, and Mustache on the client side. This question seeks a simple, single-file template solution, without using external libs.
I want to achieve this without a framework, library, etc. I'd like to put the HTML and content-list in the same .html file.
Definitely no database. It should be quick and simple to set it up within a page, without installing or configuring additional services.
Ideally, i'd like to do this without javascript, but that's not a strict requirement. If there's javascript, it should be ignorant of the fieldnames. Ideally, very short and simple. No jquery please.
you mean Template literals (Template strings) ?
const arrData =
[ { img: '/data/khang.jpg', link: 'https://khangssite.com', txt: 'Khang Le' }
, { img: '/data/sam.jpg', link: 'https://samssite.com', txt: 'Sam Smith' }
, { img: '/data/joy.jpg', link: 'https://joyssite.com', txt: 'Joy Jones' }
, { img: '/data/sue.jpg', link: 'https://suessite.com', txt: 'Sue Sneed' }
, { img: '/data/dog.jpg', link: 'https://dogssite.com', txt: 'Brown Dog' }
, { img: '/data/cat.jpg', link: 'https://catssite.com', txt: 'Black Cat' }
]
const myObj = document.querySelector('#my-div')
arrData.forEach(({ img, link, txt }) =>
{
myObj.innerHTML += `
<p>
<img src="${img}">
<br>
<a target='_blank' href="${link}">${txt}</a>
</p>`
});
<div id="my-div"></div>
This answer is a complete solution. It's exciting to edit the HTML template in codepen and watch the layout of each copy change in real time -- similar to the experience of editing a CSS class and watching the live changes.
Here's the code, followed by explanation.
HTML
<span id="template-container"></span>
<div hidden id="template-data">
IMG,, LINK,, CAPTION
https://www.referenseo.com/wp-content/uploads/2019/03/image-attractive.jpg,, khangssite.com,, Khang Le
https://i.redd.it/jeuusd992wd41.jpg,, suessite.com,, Sue Sneed
https://picsum.photos/536/354,, catssite.com,, Black Cat
</div>
<template id="art-template">
<span class="art-item">
<p>
<a href="${LINK}" target="_blank">
<img src="${IMG}" alt="" />
<br>
${CAPTION}
</a>
</p>
</span>
</template>
Javascript
window.onload = function LoadTemplate() {
// get template data.
let sRawData = document.querySelector("#template-data").innerHTML.trim();
// load header and data into arrays
const headersEnd = sRawData.indexOf("\n");
const headers = sRawData.slice(0, headersEnd).split(",,");
const aRows = sRawData.slice(headersEnd).trim().split("\n");
const data = aRows.map((element) => {
return element.split(",,");
});
// grab template and container
const templateHtml = document.querySelector("template").innerHTML;
const container = document.querySelector("#template-container");
// make html for each record
data.forEach((row) => {
let workingCopy = templateHtml;
// load current record into template
headers.forEach((header, column) => {
let value = row[column].trim();
let placeholder = `\$\{${header.trim()}\}`;
workingCopy = workingCopy.replaceAll(placeholder, value);
});
// append template to page, and loop to next record
container.innerHTML += workingCopy;
});
};
New version on github:
https://github.com/johnaweiss/HTML-Micro-Templating
Requirement
As specified in the question, this solution is intended to optimize the coding experience on the HTML side. That's the whole point of any web templating. Therefore, the JS has to work a little harder to make life easier for the HTML programmer.
The question seeks a reusable solution. Therefore, JS should be ignorant of the template, fields, and data-list. So unlike #MisterJojo's answer, the template and all data are in my HTML, not javascript. The JS code is generic.
Design
My solution is based on the <template> tag, which is intended for precisely this usage. It has various advantages, like the template isn't displayed, processed, or validated by the browser, so it has less impact on performance. Programmer doesn't have to write an explicit display:none style.
https://news.ycombinator.com/item?id=33089975
However, <template> tags are normally only intended for loading content into the layout. That's inadequate. This tool allows template variables anywhere in the HTML, including inside the tags (eg attributes like <img src).
HTML
My HTML has three blocks:
template: The HTML coder develops their desired display-structure of the output, in real HTML (not plain text). Uses <template>
data: The list of records each of which should be rendered using the same template. Uses <span> with a HIDDEN attribute.
container: The place to display all the output blocks. Uses <span>.
Template
My sample template includes 3 placeholders for data:
${LINK}
${IMG}
${CAPTION}
But of course you can use any placeholders, any number of them. I use string-literal delimiting-style (although i'm not actually using them as string-literals -- i just borrowed the delimiter style.)
Data Element
The question specifies data should be stored in HTML. It should require minimal keystrokes.
I didn't want to redundantly retype the fieldnames on every row. I didn't use slotting, JSO, Jason, or XML syntax, because those are all verbose.
https://developer.mozilla.org/en-US/docs/Web/Web_Components/Using_templates_and_slots
It's a simple delimited list. I eliminated all braces, brackets, equals, parens, colons etc.
I put the fieldname-headers only on the first row. The headers are a visual aid for the HTML developer, and a key for Javascript to know the fieldnames and order.
Record Delimiter: End-of-line
Field Delimiter: Double-commas. Seems safe, and they're easy to type. I don't expect to see double-commas in any actual data. Beware, the developer must enter a space for any empty cells, to prevent unintended double-commas. The programmer can easily use a different delimiter if they prefer, as long as they update the Javascript. You can use single-commas if you're sure there will be no embedded commas within a cell.
The data block is hidden using the hidden attribute. No CSS needed.
It's a span to ensure it takes up no room on the page.
JAVASCRIPT
Data
The data is processed by Javascript with two split statements, first on newline delimiter, then on the double-comma delimiter. That puts the whole thing into a 2D array. My JS uses trims to get rid of extra whitespace as needed.
Place-holder Substitution
Handling multiple entries requires plugging each entry into the template.
i went with simple string-replacement instead of string literals.
Multiple Templates
New version which supports multiple templates, and ability to use same template in multiple locations on same page.
https://github.com/johnaweiss/HTML-Micro-Templating
Future
Inspired by #MisterJojo, an earlier version of my solution used template literals to do the substitution. However, that was a bit more complicated and verbose, and seemed to require use of eval. So i switched to .replaceAll. Yet template-literals seems like a more appropriate method for templates, so maybe i'll revisit that.
A future version may adapt to whatever custom field-delimiter the HTML developer uses for the data block.
The dollar-curly delimiter for placeholders is a bit awkward to type. So i'm interested in finding a less awkward non-alpha delimiter that won't conflict with HTML. Considering double-brackets or braces [[NAME]]
Maybe there are simpler ways to pull the data-table into JS.
I've read components work well with <template>, but i didn't go there.
Imo, the JS committee should develop a variable-placeholder feature for <template> tags, and natively accommodate storing the data in HTML. It would be great if something like this solution was part of the rendering engine.

What are these HTML <c- g> tags? Undefined custom element?

When looking at the source code of the HTML standard there were some tags that I didn't recognise..
For example in this snippet:
<pre><code class='idl'>[<c- g>Exposed</c->=<c- n>Window</c->]
<c- b>interface</c-> <dfn id='htmlparagraphelement' data-dfn-type='interface'><c- g>HTMLParagraphElement</c-></dfn> : <a id='the-p-element:htmlelement' href='dom.html#htmlelement'><c- n>HTMLElement</c-></a> {
[<a id='the-p-element:htmlconstructor' href='dom.html#htmlconstructor'><c- g>HTMLConstructor</c-></a>] <c- g>constructor</c->();
// <a href='obsolete.html#HTMLParagraphElement-partial'>also has obsolete members</a>
};</code></pre>
From https://html.spec.whatwg.org/multipage/grouping-content.html
I thought these may be custom elements, but it doesn't look like they are defined via the custom element registry.. This is the result of interrogating the customElements object.
>>> customElements.get('c')
undefined
>>> customElements.get('c-')
undefined
Is this allowed? (I'd guess so since it's from the HTML standard, but it's still surprising to me). How would the browser know how these elements are supposed to be displayed? For example display: block vs. display: inline.
These are custom-elements (and valid HTML), generated by bikeshed's highlighter.
There is no need to define these as customElements because they don't bring any particular behavior, all they do is to ... save bandwidth.
Here is the commit excerpt:
🚨 TERRIBLE-HACK-ALERT 🚨 Switch to using <c- kt> instead of <span clas…
…s='kt'> to cut the weight of highlighting in half. Still valid HTML!
So apparently by switching from <span class="kt"> to <c- kt> (and span.kt { to c-[kt]{) they saved half of the weight induced by their highlighting.
Though as they say, it's a "terrible-hack", which still can make sense when building a tool that generates the majority of Web Standards pages, which can get very lengthy.
Regarding the default display of such custom-element, I'll quote Alohci's comment which did put it nicely:
All elements take the initial, or inherited for inherited properties, value of each CSS property until specified otherwise. So they would be display:inline
And regarding your expectation to see only best practices in the specs sources, it's better not assume so. Read the content of these pages, don't look at how they're built.
Most HTML editors don't look at the tools that will generate the pages, they write the specs in a pseudo-HTML language full of templates.
Or as it's put in the source:
<!-- Note: This file is NOT HTML, it's a proprietary language that is then post-processed into HTML. -->

Adding HTML to Word using OpenXML but unable to style the content (.Net Core)

I managed to add HTML (text only) to a Word-document following this post Add HTML String to OpenXML, using an already existing Word-file.
Unfortunately, I can't find any solution to use style from this Word-template for my newly added text. It is always "Times New Roman" size 12px although the standard style of the used template is "Arial" size 9px.
So fare I tried:
Using the ParagraphProperties as I would do for not HTML texts.
Paragraph para = body.AppendChild(new Paragraph());
Run run = para.AppendChild(new Run());
run.AppendChild(altChunk);
para.ParagraphProperties = new ParagraphProperties(new ParagraphStyleId() { Val = "berschrift2" });
Turnig MatchSource off
AltChunkProperties altChunkProperties = new AltChunkProperties();
altChunkProperties.MatchSource = new MatchSource() { Val = new OnOffValue(false) };
altChunk.AppendChild<AltChunkProperties>(altChunkProperties);
Any suggestions?
EDIT:
I found a workaround, which isn´t really a solution for my question, but works for me. I'm no longer trying to use the style from word, but adding the styles to my html before using altchunk.
Some explanation: if you look at the definition of altChunk in ISO 29500-1 17.17.2.1 and specifically in the A.1 section, the schema shows that altChunk is a EG_BlockLevelElts element and this is a peer with paragraphs (i.e. ). It is technically not correct to add as a child to run elements or even paragraph. It should be added at the body level. The fact that Word doesn't complain when adding as a run or paragraph child is unintentional and shouldn't be relied on.
As a result, what Word is doing is using the default style property for fonts to format this new content. You can try this by changing the document defaults in the styles.xml part. With match source property set to false, there isn't a way to specify the font besides document defaults.
Having said that, I think that Thomas' alternative is a better way to go.
The real solution for your question is to transform HTML into Open XML markup "yourself" rather than relying on the alternative format import parts in conjunction with w:altChunk elements. This creates a dependency on how Microsoft Word handles the import, often with little control on your side.
How do you transform HTML (or XML in general) to Open XML markup? The best way is to write so-called recursive pure functional transformations, which translate HTML elements and attributes to Open XML elements and attributes. If you have really simple HTML documents, that is not a big task. However, doing this for "arbitrary" HTML and CSS is quite a feat.
The good news is that the Open-XML-PowerTools, an Open Source library, contain functionality to transform HTML to Open XML and vice versa. Thus, I'd recommend you have a look at that library.
What worked for me and for my situation (if you don't want to go down the rather complex openxml powertools html converter root) is to add a HTML style attribute to the body section of your HTML fragment as follows:
Encoding.UTF8.GetBytes(
#$"<html><head><title></title></head><body style=""font-family: Calibri"">{ConvertUnconventionalUnicodeCharsToAscii(htmlAsString)}</body></html>");
It might be possible to dynamically derive the font family of the "normal" style embedded into the document you are updating and insert that name into the style attribute if deemed compatible.
That way, if you decide to change the base/ normal font the style of the HTML import will attempt to utilise the same font family.
Sorry if a bit off topic, I also could not get alternativeFormatImportPart.FeedData() to process "’" (code 8217) UTF-16 characters and so had to specifically replace them with "'" (code 39) in order to avoid them from being rendered as the following sequence ’

label or #html.Label ASP.net MVC 4

Newbie to ASP.net MVC 4 and trying to make sense of Razor. If I wanted to just display some text in my .cshtml page, can I use
<label class="LabelCSSTop">Introduction</label>
or should I use:
#Html.Label("STW", htmlAttributes: new { #class = "LabelCSSTop" })
Not sure if one is preferred over the other or if either is okay. If the latter emits the label tag anyway, should I just stick to the former?
Again, if I just wanted to display a text box, can I just do this:
<input id="txtName" type="text" />
or should I do this:
#Html.TextBox("txtName", "")
Is there a situation when I should use the #Html over the regular html tag?
Thanks in advance!!
In the case of your label snippet, it doesn't really matter. I would go for the simpler syntax (plain HTML).
Most helper methods also don't allow you to surround another element. This can be a consideration when choosing to use/not use one.
Strongly-Typed Equivalents
However, it's worth noting that what you use the #Html.[Element]For<T>() methods that you gain important features. Note the "For" at the end of the method name.
Example:
#Html.TextBoxFor( o => o.FirstName )
This will handle ID/Name creation based on object hierarchy (which is critical for model binding). It will also add unobtrusive validation attributes. These methods take an Expression as an argument which refers to a property within the model. The metadata of this property is obtained by the MVC framework, and as such it "knows" more about the property than its string-argument counterpart.
It also allows you to deal with UI code in a strongly-typed fashion. Visual Studio will highlight syntax errors, whereas it cannot do so with a string. Views can also be optionally compiled along with the solution, allowing for additional compile-time checks.
Other Considerations
Occasionally a HTML helper method will also perform additional tasks which are useful, such as Html.Checkbox and Html.CheckboxFor which also create a hidden field to go along with the checkbox. Another example are the URL-related methods (such as for a hyperlink) which are route-aware.
<!-- bad -->
my link
<!-- good -->
#Html.ActionLink( "my link", "foo", "bar", new{ id=123 } )
<!-- also fine (perhaps you want to wrap something with the anchor) -->
<span>my link</span>
There is a slight performance benefit to using plain HTML versus code which must be executed whenever the view is rendered, although this should not be the deciding factor.
Depends on what your are doing.
If you have SPA (Single-Page Application) the you can use:
<input id="txtName" type="text" />
Otherwise using Html helpers is recommended, to get your controls bound with your model.
If you want to just display some text in your .cshtml page, I do not recommend #Html.Label and also not to use the html label as well. The element represents a caption in a user interface. and you'll see that in the case of #Html.Label, a for attribute is added, referring to the id of a, possibly non-existent, element. The value of this attribute is the value of the model field, in which non-alphanumerics are replaced by underscores.
You should use #Html.Display or #Html.DisplayFor, possibly wrapped in some plain html elements line span or p.
The helpers are there mainly to help you display labels, form inputs, etc for the strongly typed properties of your model. By using the helpers and Visual Studio Intellisense, you can greatly reduce the number of typos that you could make when generating a web page.
With that said, you can continue to create your elements manually for both properties of your view model or items that you want to display that are not part of your view model.
When it comes to labels, I would say it's up to you what you prefer. Some examples when it can be useful with HTML helper tags are, for instance
When dealing with hyperlinks, since the HTML helper simplifies routing
When you bind to your model, using #Html.LabelFor, #Html.TextBoxFor, etc
When you use the #Html.EditorFor, as you can assign specific behavior och looks in a editor view
#html.label and #html.textbox are use when you want bind it to your model in a easy way...which cannot be achieve by input etc. in one line

Search underlined link in watir

I am trying to search/select a link in a page that is underlined, while others are not. The source is something like this:
<a href="someurl1">
<b>
<u>Some ulined text</u>
</b>
<u></u>
</a>
<br>
Other link text
<br>
Another Link text
<br>
I tried something like
link = browser.link(:u?, true)
link.exists?
I get the following errors
TypeError: expected one of [String, Regexp], got true:TrueClass
from /usr/lib/ruby/gems/1.8/gems/watir-webdriver-0.2.4/lib/watir-webdriver/locators/element_locator.rb:152:in `check_type'
from /usr/lib/ruby/gems/1.8/gems/watir-webdriver-0.2.4/lib/watir-webdriver/locators/element_locator.rb:189:in `normalized_selector'
from /usr/lib/ruby/gems/1.8/gems/watir-webdriver-0.2.4/lib/watir-webdriver/locators/element_locator.rb:188:in `each'
from /usr/lib/ruby/gems/1.8/gems/watir-webdriver-0.2.4/lib/watir-webdriver/locators/element_locator.rb:188:in `normalized_selector'
from /usr/lib/ruby/gems/1.8/gems/watir-webdriver-0.2.4/lib/watir-webdriver/locators/element_locator.rb:76:in `find_first_by_multiple'
from /usr/lib/ruby/gems/1.8/gems/watir-webdriver-0.2.4/lib/watir-webdriver/locators/element_locator.rb:33:in `locate'
from /usr/lib/ruby/gems/1.8/gems/watir-webdriver-0.2.4/lib/watir-webdriver/elements/element.rb:260:in `locate'
from /usr/lib/ruby/gems/1.8/gems/watir-webdriver-0.2.4/lib/watir-webdriver/elements/element.rb:247:in `assert_exists'
from /usr/lib/ruby/gems/1.8/gems/watir-webdriver-0.2.4/lib/watir-webdriver/elements/element.rb:31:in `exist?'
Edit:
So I am actually using this for screen scraping rather than testing. That may explain the reasons why watir does not support this directly since CSS and other better practices make sense for testing and when you the HTML development and testing go hand in hand. Hoserver from a scraping perspective, the text formatting is what the user sees, and searching underlined, bold links etc. make sense for scraping.
I've never seen that kind of attribute used in a test case before. I also haven't seen any code support for it. You may have to roll your own. Here is an example borrowed from Zeljko
def hasUnderlined(browser)
s = false
browser.links.each do |l|
if l.html.downcase.match /\<u\>*\<\/u\>/
s = true
end
end
end
def getUnderlined(browser)
browser.links.each do |l|
if l.html.downcase.match /\<u\>*\<\/u\>/
return l
end
end
end
I don't think what you want is possible directly because the underline is not an attribute of the link tag, but a formatting tag that apples to just the text in the link.
However, in modern web pages, formatting is often controlled by a combination of CSS and attributes such as class names, which ARE something you could specify when identifying a link. So IMHO your best bet here might be to talk a little with your developers about how they are coding the site and see if they are perhaps open to increasing the testability of their code by using slightly more modern techniques for controlling what links are underlined, such as say using CSS and basing the underlining on a class name. (There's a lot of other good reasons to use CSS for controlling formatting instead of embedding it directly in the HTML, but unless your guys are fresh off the html-banana-boat so to speak, they should not need to be taught why using CSS is a good thing)
That would let you search for a link according to the class attribute that was being used to cause CSS to underline the text
If your developers are not open to such an approach to make their code more testable, then I think your only option is going to be to create your own ruby code for this and modify your copy of water (see #Dave's answer), and then be prepared to maintain that custom patch any time you update watir etc.
So, the only thing you know about a link is that it is underlined?
I thought this would do it (using the latest watir-webdriver gem):
browser.link.u.click
But I got this:
NoMethodError: undefined method `u' for #<Watir::Anchor:0x00000100cbb3c0>
Jari (watir-webdriver developer) said he thinks u tag is not in HTML spec.
By the way, this works:
browser.link.b.click
Jari suggested trying xpath, but I thought css selectors would be nicer to read. Here it is:
browser.element(:css => "a b u").click