----- Original Message -----
From: "Paul Novitski" <paul@xxxxxxxxxxxxxxxxxxx>
At 11/28/2006 05:05 AM, Satyam wrote:
May I invite you to check http://satyam.com.ar/pht/? This is a project I
started some time ago to help me produced HTML in a more clean and
efficient way. Usually, producing good HTML involves running a sample
output through some HTML validator or looking at the HTML and checking it
out by hand, which, of course, requires good formatting to make it
understandable. In the case of too dynamic HTML (meaning, output can
vary widely) it is impossible to produce enough samples of all the
possible outputs to get them checked out.
So, my idea was to embed HTML into the language itself so that the final
output could be checked at 'compile time' and even before, using the
standard tools provided by the IDE (even just matching braces goes a long
way into checking for missmatched HTML tags).
Satyam,
That's an impressive bit of work, congratulations.
It's interesting to see someone spend such energy merging PHP logic with
HTML. I've gone in the opposite direction, separating the two as much as
possible. My own CMS merges content with HTML based on CSS-style
selectors so that the logic layer of my applications doesn't need to
know -- or contain -- the full details of the markup. I find this a
natural and agreeable extension of the move to separate HTML markup from
CSS presentation and JavaScript behavior.
It's interesting to note that for all your effort to generate good, clean
HTML, you're still able to generate a div nested inside a table:
<table for ( $i=1; $i < 10; $i++) {
<div {
&style = ($i & 1?"odd":"even");
<tr {
Ouch! According to the spec, this is an illegal structure:
http://www.w3.org/TR/html4/struct/tables.html#h-11.2.1
Your pre-compiler has ensured that all your tags are well-formed, but it
doesn't ensure that you've followed the rules of correct markup. Perhaps
a future iteration of your software will incorporate more HTML structural
rules and will give you precompiler errors in such cases.
That would be something to be dealt with in the second part of the project,
that of validating the HTML output against the DTD. Actually, the
pre-compiler is meant for generic XML, thus it does not know about divs and
tables, it just knows about how to output properly formatted XML.
You write:
So, we have two well structured languages, one is procedural (any flavor
of C, JavaScript, PHP, Perl, etc.), the other descriptive. Their blocks
are quite compatible: to start with, they nest nicely within each other.
If an XML block is contained within an if() block, it has to be completely
within it, the boundaries of their blocks should not overlap.
Help me understand the relevance of this statement. A very common pattern
in a mixed logic/HTML script goes like this:
echo '<ul>';
foreach (array as item)
{
echo '<li>item</li>';
}
echo '</ul>';
In these cases the boundaries of the HTML block do in fact overlap the
boundaries of the foreach() logic block.
Perhaps 'overlap' is not the right word, I mean that one is fully contained
within the other. The braces of the foreach are fully enclosed by the ul
and the li tag is fully within the braces.
And, of course, I would appreciate any comment on the project, EXCEPT that
you use template engines and that you do not generate HTML directly. I've
heard that and it is completely missing the point so, please, spare me
that one. At one point or another plain HTML has to be generated.
Unless I'm missing the boat, it seems to me that the primary advantage of
your precompiler is that it enables you to close HTML tags simply by
closing braces, a convention policed by your pre-compiler and the PHP
interpreter itself, so that you'll get pre-compiler or interpreter errors
for incorrect closure instead of waiting for the W3C validator to check
your work. I don't mean to minimize the significance of your
accomplishment, but personally I don't find generating accurate markup to
be a great problem. I'm a careful hand-coder, and true to the topic of
this thread I find that neatly-indented HTML helps me validate my own
markup. Online validators help me catch any errors I miss.
Without the second part, validating against the DTD, yes, it doesn't do much
more than you say.
What I find to be a much greater problem is the human readability of logic
code when HTML is mixed throughout. Your innovation is helpful here, as
you're nearly making HTML tags into PHP key words, eliminating some of the
literal quoting that makes PHP+HTML so tiresome. However, even with your
pre-compiler the messy quotes are still there on the attribute level.
The value of the attribute is any valid PHP expression and if they are
literal strings there will be quotes, but then, you can also use heredoc if,
for example, you are putting some JavaScript code into an event. If the
value is a numeric value, there is no need for any quotes at all. In the
end, the value of an attribute is any valid PHP expression and it follows
PHP rules.
And, stepping back, you're perpetuating the embedding of markup with logic
so that it will still take a PHP programmer to modify the markup of one of
your pages. Do you not see the advantage in separating the two layers?
Yes, I do, and I would recommend using templates or similar tools to provide
for separation of code and markup, but sometimes there are reasons not to do
so, for example, web services. Though the examples are in HTML so I don't
have to explain the semantics of an arbitrary markup language, it is
actually meant for XML where there is no presentation layer at all. Lots of
packages, (Wordpress, many picture galleries and CMSs), don't do much
separation of layers at the HTML / PHP interface. They do have separate
modules for presentation and for program logic and also data handling, but
the presentation layer is not plain HTML with template tags but more of a
mix of PHP /HTML so this would still work for those.
Again, in spite of this criticism I'm impressed with your effort. Good
work!
I appreciate your comments and, I admit, my main purpose in doing this was
to learn how to do it. I am an engineer and when I studied, a couple of
semesters of Fortran IV was all I got (and punching cards at that, yes, I am
that old), all the rest was self-taught so I wanted to go deeper into some
aspects of computer science such as compilers (there is also a PHP grammar
for JavaCC which I made earlier in the process).
In fact, my original idea was some sort of embedded SQL as it exists for C,
but I know it does not work quite Ok, in fact, it has been there for quite
some time and it doesn't catch up. SQL is such a different kind of beast
that it is hard to make it compatible. SQL cursors and error handling are
concepts which are hard to blend into a procedural language so I believe it
is better to handle SQL through functions where it is clearly separate from
the language calling them. Thus, I thought, we have three main languages
here, HTML, PHP and SQL. I know PHP and SQL don't mix well, how about the
other end? That's when I started to think about this pre-compiler and found
it to be a pretty logical mix.
Cheers and thanks again for your comments.
Satyam
Regards,
Paul
--
PHP General Mailing List (http://www.php.net/)
To unsubscribe, visit: http://www.php.net/unsub.php
