The following articles were found with tag technicalarticles:
Context Switching Layouts with the Zend Framework
Wed Jul 23, 2008 at 13.31
If you have not been hiding under a rock for the past few years, you realize that there are many different platforms and mediums to which a web site or application might wish to publish. The guys at Zend came up with a tool that can be used to facilitate multiple device view rendering in a semi-transparent way. This is known as Context Switching in the Zend Framework and is accomplished using the ContextSwitch Action Helper.
One hurdle I came across was figuring out a simple way to provide a consistent user experience through what is known as a two-stage view rendering. This is done using Zend_Layout. However, layouts written in standard XHTML won't really work well for our mobile platform, and likewise for XHTML-MP in a browser on a PC. Enter in context switching. First we are going to set up our context switching inside the index controller:
{
public function init()
{
$context = $this->_helper->getHelper('contextSwitch');
$context->addContext('mobile', array('suffix' => 'mobile'))
->setAutoDisableLayout(false)
->addActionContext('index', 'mobile')
->initContext();
}
public function indexAction()
{
//do something here, it's unimportant for this post
}
}
This will allow us to have a mobile view for urls such as http://example.com/index/index/format/mobile (NOTE: this should be prettier, but routing is outside the scope of this post). The key to take from this short code sample is the setAutoDisableLayout(false) bit.
Now, assuming we have created a view script called index.mobile.phtml we will now see the contents of the mobile view script dumped inside the default layout script. This isn't exactly what we wanted, believe it or not. Well, not really what we wanted but it's pretty close. We just need to tell the Layout manager that we want a contextual layout as well as a contextual view script. We do that using the layout helper:
{
$context = $this->_helper->getHelper('contextSwitch');
$context->addContext('mobile', array('suffix' => 'mobile'))
->setAutoDisableLayout(false)
->addActionContext('index', 'mobile')
->initContext();
$layout = $this->_helper->getHelper('layout');
$currentContext = $context->getCurrentContext();
if($currentContext)
{
$layout->setLayout($layout->getLayout() . "." . $currentContext);
}
}
After we have a new layout script called layout.mobile.phtml we should be good to go.
References
Zend Framework: ContextSwitch Action Helper
Zend Framework: Layout Configuration Options (see: getLayout())
- Tagged:
- technicalarticles
- php
- Zend Framework
With Continuous Integration, Who Needs a Manger?
Wed Jun 11, 2008 at 14.12
For anyone who is not familiar continuous integration (CI) is basically the practice designed to leverage automated testing and to continuously (usually daily, sometimes more) integrate changes from multiple developers using a build tool. To be successful and useful the code being integrated should be fairly well tested; meaning, something Martin Fowler refers to as "self testing code." However, one could undertake the practice and slowly introduce the comprehensive test suite as the days go by. (Notice I didn't leave any room for the lazy route of not having the test suite?).
The role played by the CI server is to on a set interval, in reaction to code updates, or on a combination thereof run a build process that incorporates testing to ensure the code checked in by each developer meshes with the overall project and does not violate the rules set forth in the unit tests. By doing this after each commit, the server verifies the stability of the code and basically stamps the product as "working". Does this remind you of what your manager is supposed to do? Is (s)he routinely (daily) checking on your and your colleagues progress and making sure everyone is on the same page? Does (s)he verify that your project is meeting the original goals of the project? Can he/she at any given time produce metrics to show the amount of code that is successfully passing tests, the amount of code that is untested, or the number of times checked-in code causes a problem in the build? Chances are, probably not. But with the help of this type of tool the answer will quickly become yes.
What I've found in my time working with (under) a CI server is I am constantly aware of what my development efforts produce. I know that once every two hours a daemon on my machine will run a script that checks my source control repository and compares the current working copy to the copy in version control. If the copy under version control is newer than the working copy the CI server runs a build script which runs the unit testing complete with a generated coverage report, runs a script to create documentation from code commenting, and copies the newly developed files to the development webserver where they can be viewed as if in the production environment. I know that I missed the mark and only have 83.0 percent of my code base covered by unit testing. I know that I've had 37 successful and 4 failed builds which shows me that 90% of the time the code I've checked in to the repository is well-tested. The other 10% of the time I must be in la-la land.
In various positions I've held meeting frequencies have varied from monthly to weekly, but never have I had the pleasure of a bi-hourly accountability like I do with my CI server. I say all this to tell you that not only is CI a great tool to help take care of some otherwise time-consuming and mundane tasks but it is also a wonderful driving force to create better development habits and more stable code.
- Tagged:
- technicalarticles
- php
Deriving values for a non-existent table
Tue Oct 16, 2007 at 16.06
This situation might arise in rare cases but if it does there is a simple solution that should get you on your way. Here is the problem: You have two tables that are related by a column in each of them. However, there may or may not be a record in either of the tables for any given key. It is a bit tricky to explain in words, so perhaps an example is in order. For whatever reason you are working with the following tables
game_id int,
venue_id int,
game_date datetime,
total_points int,
team_id int,
primary key(game_id, venue_id, game_date),
unique ven_date_uniq(venue_id, game_date)
);
insert into league_games values
(1, 1, '2007-01-01 12:00:00', 89, 1),
(2, 1, '2007-01-01 16:30:00', 72, 2),
(3, 1, '2007-01-02 11:00:00', 103, 7),
(4, 1, '2007-01-02 19:30:00', 97, 10),
(5, 2, '2007-01-08 16:00:00', 65, 4),
(6, 2, '2007-01-08 19:00:00', 78, 2)
(17, 2, '2007-01-09 17:30:00', 102, 12),
(18, 2, '2007-01-09 19:00:00', 58, 13),
(19, 3, '2007-01-19 14:30:00', 78, 17),
(20, 3, '2007-01-19 16:00:00', 87, 22),
(21, 3, '2007-01-19 18:00:00', 109, 27);
create table tournament_games(
tournament_id int,
game_id int,
game_date datetime,
total_points int,
team_id int,
primary key(tournament_id, game_id),
unique team_date_uniq(team_id, game_date)
);
insert into tournament_games values
(1, 7, '2007-03-07 09:30:00', 88, 1),
(1, 8, '2007-03-07 12:00:00', 76, 3),
(1, 9, '2007-03-07 14:30:00', 92, 5),
(1, 10, '2007-03-07 17:00:00', 80, 7),
(1, 11, '2007-03-07 19:30:00', 109, 9),
(1, 12, '2007-03-08 09:30:00', 97, 2),
(1, 13, '2007-03-08 12:00:00', 56, 4),
(1, 14, '2007-03-08 14:30:00', 85, 6),
(1, 15, '2007-03-08 17:00:00', 105, 8),
(1, 16, '2007-03-08 19:30:00', 93, 10);
From these tables we are trying to calculate the total number of points accumulated by each team in both tournament and league play for the Spring 2007 (January - May 2007) season. It is important to note that we do not have a teams table to join against, which is the crux of the problem. However before we jump into a solution we need to discuss another fact. While it may seem a little counter-intuitive that a team would play in the tournament but not in the leauge, our data supports this situation. Also a team may play in the league but not in the tournament. This adds a dimension to our problem because you must account for the missing values (which in an OUTER JOIN will produce NULLs) in your query. This causes problems because the sum of an integer and NULL is not defined, so it is NULL. But NULL isn't a number so the total points cannot be NULL. To get around this problem we will use the COALESCE function which takes the first non-NULL value in its argument list as the return value. Let's have a look at our first query attempt
, sum(coalesce(l.total_points, 0) + coalesce(t.total_points,0)) total_points
from league_games l
left outer join tournament_games t
on l.team_id = t.team_id
and t.game_date > '2006-12-31 23:59:59'
and t.game_date < '2007-05-01 00:00:00'
where l.game_date > '2006-12-31 23:59:59'
and l.game_date < '2007-05-01 00:00:00'
group by l.team_id
Now this works fine provided there are no teams in the tournament that are not also in the league play. Given the connotation of league and tournament play, it doesn't make sense for this situation to arise, but humor me. We are, given the data above, missing point totals for teams 3, 5, 9, 6, and 8. To get these values we can reverse the order of the tables in query but then we will be missing totals for teams 4, 12, 13, 17, 22, and 27. The only way is to add a third table to the query that gives the valid teams. In an ideal situation we would have a table, teams or even teams_seasons, that would provide the necessary information, but we are not so lucky in our example. To get this data we will derive our teams table in the query using a union. Here is the query to derive the teams table
from league_games l
where l.game_date > '2006-12-31 23:59:59'
and l.game_date < '2007-05-01 00:00:00'
union
select l.team_id
from tournament_games l
where l.game_date > '2006-12-31 23:59:59'
and l.game_date < '2007-05-01 00:00:00'
order by team_id
This produces a list of team_ids that appeared in the Spring 2007 season which we can join against both games tables to get our final result. You might be tempted to just join against this table and forget about the date range for the games tables, but that would be a mistake. While it is possible that a given team only played during one season, it is more likely that they played in multiple seasons and not including the conditions on the games tables would skew your results to total points over all time and not only in the given season. Putting it all together we have our final solution
, sum(coalesce(l.total_points, 0) + coalesce(t.total_points,0)) total_points
from (
select l.team_id
from league_games l
where l.game_date > '2006-12-31 23:59:59'
and l.game_date < '2007-05-01 00:00:00'
union
select l.team_id
from tournament_games l
where l.game_date > '2006-12-31 23:59:59'
and l.game_date < '2007-05-01 00:00:00'
order by team_id
) teams
left outer join league_games l
on teams.team_id = l.team_id
and l.game_date > '2006-12-31 23:59:59'
and l.game_date < '2007-05-01 00:00:00'
left outer join tournament_games t
on teams.team_id = t.team_id
and t.game_date > '2006-12-31 23:59:59'
and t.game_date < '2007-05-01 00:00:00'
group by l.team_id
order by teams.team_id
Which correctly answers our query and is an elegant solution to the presented problem.
- Tagged:
- technicalarticles
- database
- mysql
Bad Practice: DB Queries inside loops
Sat Sep 29, 2007 at 03.37
This is something I see day in and day out, and unfortunately is a deep gash into the good reputation of otherwise awesome PHP coders. If you ever find yourself putting a database query inside a loop, stop and study the code for a long moment. Chances are this is unnecessary.
Generally speaking the only acceptable time for a query to be inside a loop is when it is contained in a script that will only be used once and that is meant to do some sort of data transfer/transformation (which can probably be better accomplished using a pure SQL solution, but that is a battle for another day.)
So let's examine one of the more common occurrences of this problem:
, SUM(points) total_points
FROM user_points
GROUP BY user_id
ORDER BY total_points DESC";
$query_1 = mysql_query($sql_1) or die(mysql_error() . "<br />" . $sql_1);
while($row_1 = mysql_fetch_assoc( $query_1 )) {
$sql_2 = "SELECT first_name
, last_name
FROM users
WHERE user_id = " . $row_1['user_id'];
$query_2 = mysql_query( $sql_2 ) or die( mysql_error() . "<br />" . $sql_2);
//more code follows
}
If you recognize this code and are thinking, "Hey man, what's wrong with that? That's how I always do it!" then I am extremely happy that you are reading this post.
The problem is that we really don't know how many users there are - and even if we did this is still a bad solution - in the user_points table and thus we are repeating the query to get the user's name an unknown number of times (N times.) This is a resource hog of a script to do something that is simple to accomplish using the proper query.
In this case, had we known about and employed an INNER JOIN, we will eliminate the subsequent calls to the database and retrieved all the information we need in one trip. This terminates the script's connection to the database quicker and allows another connection to be established by a different script. This is a good thing, trust me. If you have ever had to figure out what is causing your server to crawl you will know that I speak the truth.
So how do we solve the problem? As stated above we take that query and get the information using an INNER JOIN. Here is our new query
, u.first_name
, u.last_name
, SUM( up.points ) total_points
FROM user_points up
INNER JOIN users u
ON up.user_id = u.user_id
GROUP BY user_id
ORDER BY total_points
If any of that looks foreign to you, it's time you brushed up on your SQL. I suggest reading a few tutorials on JOINs.
- Tagged:
- technicalarticles
- phpalized
- database
- mysql
- php
Debugging MySQL Queries in PHP
Fri Sep 28, 2007 at 23.21
Over at what is becoming my second home a frequently asked question ultimately results in the original poster needed some guidance in debugging interaction between their PHP scripts and their database, namely MySQL. I hope to uncover some best practices for making it easier to figure out what is going on with your script.
This first snippet is a commonly occurring one and will be known as "the hard way" in terms of finding the cause of your incorrect results quickly.
while($row = @mysql_fetch_assoc($result)) {
echo $row['user_id'] . " " . $row['username'];
}
This is the hard way because we have suppressed any errors using the @ operator and assumed that there will always be results. If something goes wrong, we might be surprised, or even worse just assume that there aren't any results that match our query. The query from above is rather simple and will probably not produce an error, but once you get into more involved queries you might have a mistake or two. Which brings me to my first point: you should create your queries outside of PHP (e.g. on the command line or in a product such as MySQL Query Browser or PhpMyAdmin) and ensure that they are working before adding them to the equation in your script.
Once you have your query working correctly directly against the database add it to your script with a few additions to the above code. First we will utilize the native function mysql_error to return the error from the MySQL engine and use it to show us exactly what went wrong. This coupled with the display of the query that was actually sent to the engine will prove extremely helpful in the future. This is how a query snippet might look with some more debugging features
$result = mysql_query($sql) or die("MySQL error: " . mysql_error() . " generated by " . $sql);
In the event of a failed query, we can now see the error generated by the db engine and the actual query that caused the error. This however is not something you should find in a production environment. The reason is outside the scope of this post, but involves security by not hinting at database structure.
When handling the utilization of result sets the result set should first be verified. We did not do this in our first example we just assumed that there were rows and that the query was successful. That can also be handled better using the following snippet
echo "Database query returned an error.";
}
elseif( mysql_num_rows( $result ) == 0 ) {
echo "There were no results returned.";
}
else {
while($row = mysql_fetch_assoc( $result) ) {
echo $row['user_id'] . " " . $row['user_name'];
}
}
Naturally you will probably want to handle the cases a little differently (i.e. not just echo out those simple statements) but you get the general idea.
- Tagged:
- technicalarticles
- phpalized
- database
- mysql
- php
Functional minimum width in IE6
Fri Sep 28, 2007 at 06.46
I learned a new fun trick to play on Internet Explorer 6 with this design. There was not much information on it, so it might just be common knowledge amongst the more practiced designers out there, but to me it is new and interesting and I thought I would share it with you.
The situation arose with coding the blog. Since I want to be able to share images that may or may not be hosted on my site and I am against the idea of manually resizing (or downloading the image and resizing it dynamically) I wanted to use a CSS rule to scale images that would break outside of the container down to a size that would fit inside the container. This is farily easy to accomplish in FireFox with a min-width, but IE doesn't play so nice. The solution I found was to use an expression within the stylesheet to get the image's width and use it as the width for the style rule if it is less than the desired width or the desired width if it is greater.
The rule for use in FireFox would be
and in the IE stylesheet the rule would be
Viola you have a functional (as of now) implementation of a minimum width in IE6.
- Tagged:
- css
- design
- technicalarticles