code

Simple PHP MVC

One of the much talked about patterns of the moment is the model-view-controller or model-view-presenter or any of the undoubtedly thousands of variants thereof. Here’s a deliberately really simple one to illustrate the pattern. If you can read it and understand it I hope its useful to you.

Interestingly a vague attempt at inversion of control, another design pattern which IMHO is more of a guide line, is in play here. It is done specifically to allow good component separation and useful test patterns. Hence as long as you have a class that fits the dependency obligations you can use it with the ones given here.

Now for the seriously crazy bit, the code on this page has not been tested to work. Its just here to give you some idea of what the model-view-controller pattern and how it could be implemented.

mvc.controller.php

Overview:

At the heart of each variant implementation of the model view controller pattern is the controller. A controller basically directs requests to the appropriate model or view processing functionality. In the case below it only makes the distinction between the two types and not the instances of functionality.

Some controllers combine a lot more but its not actually necessary and generally ends up making the controller very configuration and platform specific. By keeping configuration out of the controller we make it as general as it can be.

Notes:

Note that there is no exception handling here either. It is not wanted. Consider for a moment that if an exception was not caught, there would be no response to the client, which is a security requirement in many instances. However it there would be a log to trace to the point of failure because of the log entry call.

If exception handling were introduced it would compromise the generic nature of the controller as the controller would need to know what type of response to send to the client. For example if the error was thrown before the content of the request could be analysed and what data would you respond to the request with? HTML? XML? A media stream? A GIF? A JPG? The rule of thumb here is only respond if you know what language to talk in, if you don’t know the requested format you can’t even use clever configuration to determine how respond to a request.

Documented dependencies:

  • class ‘Model’ : method ‘process_request’ : parameters ‘request to process’ : returns ‘new request for either model processing or view generation’
  • class ‘View’ : method ‘process_request’ : parameters ‘request to process’ : returns ‘a view generated from the accumulated data in the request’

Undocumented dependencies:

  • class ‘Log’ : method ‘writestr’ : parameters ‘string to log’, ‘name of log level to record string at’ : returns ‘nothing’
  • class ‘Log’ : method ‘writereq’ : parameters ‘request to log’, ‘name of log level to record string at’ : returns ‘nothing’
  • class ‘Request’ : method ‘is_a_model_request’ : parameters ‘none’ : returns ‘true if the request is to be processed by the model’

Listing:


<?php

Log::getInstance()->writestr( "file:".__FILE__.":loaded", "debug" );

class Controller {

  private $model;

  private $view;

 /*

  * build this controller object

  */

  public function __Construct ( $model, $view ) {

    $this->model = $model;

    $this->view = $view;

  }

 /*

  * process a request

  */

  public function process_request ( $request ) {

    Log::getInstance()->writestr( 

      "class:".__CLASS__.":process_request", "debug" );

    Log::getInstance()->writereq( $request, "debug" );

   /*

    * first process the model request

    */

    $new_request = $this->model->process_request( $request );

   /*

    * if this request results in another model request

    */

    if ( $new_request->is_a_model_request( ) )

      return $this->process_request( $new_request );

   /*

    * otherwise generate view

    */ 

    return $this->view->process_request( $new_request );

  }

}

?>

mvc.model.php

Overview:

A model is responsible for the production of data for the view. Essentially it contains access points to the various methods of the underlying application called actions. In this case it simply has a list of actions that it locates and then executes the action sending the results back to the caller.

Notes:

As with the controller (see above) the model does little more than the basics leaving other classes that know what they are doing to handle the specifics of content. In line with this the model simply logs progress and fails silently.

Undocumented dependencies:

  • class ‘Log’ : method ‘writestr’ : parameters ‘string to log’, ‘name of log level to record string at’ : returns ‘nothing’
  • class ‘Log’ : method ‘writereq’ : parameters ‘request to log’, ‘name of log level to record string at’ : returns ‘nothing’
  • class ‘Actions’ : method ‘find_requested_action’ : parameters ‘request to locate action for’ : returns ‘an Action object’
  • class ‘Action’ : method ‘process_request’ : parameters ‘request to process’ : returns ‘new request for either model processing or view generation’

Listing:


<?php

Log::getInstance()->writestr( "file:".__FILE__.":loaded", "debug" );

class Model {

  private $actions;

 /*

  * build this model object

  */

  public function __Construct ( $actions ) {

    $this->actions = $actions;

  }

 /*

  * process a request

  */

  public function process_request ( $request ) {

    Log::getInstance()->writestr( 

      "class:".__CLASS__.":process_request", "debug" );

    Log::getInstance()->writereq( $request, "debug" );

   /*

    * find the action from the list of actions

    */

    $action = $this->actions->find_requested_action( $request );

   /*

    * run the action and return the resulting request

    */

    return $action->process_request( $request );

  }

}

?>

mvc.view.php

Overview:

A view is called to format the data retrieved from the model into the format that the original client requested. Essentially it contains access points to the various view methods of the underlying application. In this case it simply has a list of views that it locates and then executes sending the results back to the caller.

Notes:

Now those sharp of eye will realise something; this looks pretty much like the model class above, and you’d be right. Apart from a a simple renaming of a few items the code is the same. In practice both the above model class and this view class would be implemented as a single class and simply configured differently. They are only provided in their separate formats here to highlight the model-view-controller form.

Undocumented dependencies:

  • class ‘Log’ : method ‘writestr’ : parameters ‘string to log’, ‘name of log level to record string at’ : returns ‘nothing’
  • class ‘Log’ : method ‘writereq’ : parameters ‘request to log’, ‘name of log level to record string at’ : returns ‘nothing’
  • class ‘Views’ : method ‘find_requested_view’ : parameters ‘request to locate view for’ : returns ‘an View object’
  • class ‘View’ : method ‘process_request’ : parameters ‘request to process’ : returns ‘the final generated view’

Listing:


<?php

Log::getInstance()->writestr( "file:".__FILE__.":loaded", "debug" );

class View {

  private $views;

 /*

  * build this view object

  */

  public function __Construct ( $views ) {

    $this->views = $views;

  }

 /*

  * process a request

  */

  public function process_request ( $request ) {

    Log::getInstance()->writestr( 

      "class:".__CLASS__.":process_request", "debug" );

    Log::getInstance()->writereq( $request, "debug" );

   /*

    * find the view from the list of views

    */

    $view = $this->views->find_requested_view( $request );

   /*

    * process the request and return the resulting view

    */

    return $view->process_request( $request );

  }

}

?>

LDAP vs RDBMS is the war over?

Perhaps the question is better put, ‘Did the war even happen?’ What war you might well ask; well when different ways of looking at potentially the same data are viewed then there are bound to be zealots on either side fanning the flames of FUD on the opposition and extolling the vitues of their chosen creed. However although both RDBMS and LDAP are commonly used to host the same data for applications there appears to be little controversy.

Mostly, I think, there is no conflict because in general LDAP is not even an option for the vast masses of application developers out there. Especially as the bulk of application development appears to be done facing the Internet. Normally hosting environments provide a choice of RDBMS or RDBMS. LDAP is just too difficult a concept, and that’s probably the first reason that it has fallen out of contention for storing the masses of organisational data out there. Yet it is probably the best candidate that I know of for holding operational control of that data.

To illustrate this conceptual difficulty lets consider the contention between software developer and database architecht perspectives. Typically the developer does not really understand the various tools that the RDBMS offers to manage data. The database expert loaths the way that developers start implementing relational contraints in code instead of applying the appropriate design concepts of foriegn keys, triggers, stored procedures etc. I’ve seen this many times. Often I’ve seen the opposite where database folk have decided to embed code into their domain. To really get this right each party needs to understand the value that the other brings to the table. But I continually read laments from either side criticising the way that this or that should really be done over here and not there.

Which brings us into LDAP. This underused gem is typically used by the development community simply as a convienient pre-built authentication tool. Even when LDAP is used a lot of information is then stored in an RDBMS that it really does not do well; or at least as well as LDAP. I’ll spare you the history of why this is so and how it came to be but the fundamental difference between the two is that RDBMS generally is a collection of flat file tables that are related by loose rules; whereas the LDAP server is a tightly coupled hierarchy of objects (called the Directory Information Tree – DIT) similar in nature to that of other concepts like XML.

LDAP servers are generally built for high speed retrieval and mass replication focussed on the enterprise, where RDBMS is a general ledger store house, and pretty primative too. The kind of data then that you should store in the LDAP directory is structural data. Like say configuration data for a telephone system or contact information for a customer relationship management (CRM) type of application. Even the configuration and preferences information of WordPress for this blog should really be stored in an LDAP server. But its not.

If you want to know if the data that you are looking at is suited to LDAP then consider.

  • Is the data dynamic or relatively static?
  • Does the data need to be distributed?
  • Can the data be used by more than one application?
  • Is the data multi-valued?
  • Can your data or application take advantage of a hierarchical relationship?
  • Do you need flexible security options?
  • Do you need single sign-on?
  • Do you need distributed or delegated administration capabilities?

If you can answer yes to some or all of these questions, then directories and directory-based applications would likely be useful to your application or project.

So why did I write this post? Basically it is a pointer into the world of LDAP for those who’ve not thought of it. A start. If you really want a better introduction go here. Have fun.


Driven to Reinvent

Its nice to have a standards based platform and in that vein C certainly has progressed a long way. But you still get the odd system that isn’t either set correctly up by default with my favorite extensions; namely all things GNU! Yes I admit to liking non-standard well coded extensions, in fact who would work without them these days except under a very specific set of requirements or academic interest. Things just would not get done so quickly without them.

So I got a Mac. Yep its sooo cool that when my 5 year old daughter walked in the room and saw it for the first time she exclaimed ‘Daddy! That’s sooo cool!’ Kind of an ego boost for someone who gave up on cool and cool things 20 years ago. However cool comes at a price, its got great development tools but I want cross platform development to run on my Linux boxen too. Long story short, it doesn’t have the standard GNU extensions so I had two choices, fart around with the environment or rewrite one or two functions myself.

So here’s the goods. Now because I don’t want people to just copy and paste my material without thinking about it (I don’t mind the copying I just like people to put in some effort and think) I’ve copied an earlier version of the functions which works for some situations but not for others. Its close tot he real deal but there are about 5 significant issues which would make it dangerous to use this code without doing something about them.

Here it is, have fun, have a play! (Note: the whole thing is under GPL 3, comments from GNU)


/*

ssize_t getdelim (char **lineptr, size_t *n, int delimiter, FILE *stream)  

This function is like getline except that the character which tells it to stop 

reading is not necessarily newline. The argument delimiter specifies the 

delimiter character; getdelim keeps reading until it sees that character (or end 

of file).

The text is stored in lineptr, including the delimiter character and a 

terminating null. Like getline, getdelim makes lineptr bigger if it isn't big 

enough.

getline is in fact implemented in terms of getdelim

*/

ssize_t getdelim( char **lineptr, size_t *n, int delimiter, FILE *stream ) 

{

	/* setup the environment */

	int c = getc( stream );

	long i = 0;

	char *a = NULL;

	/* count the characters needed for the buffer */

	while ( c != delimiter && c != EOF ) {

		c = getc( stream );

		i++;

	}

	/* test for and arrange the buffer size */

	if ( i == 0 ) 

		return -1;

	if ( fseek( stream, -i, SEEK_CUR ) ) 

		return -2;

	if ( *n < i + 1 ) {

		a = ( char * )realloc( *lineptr, i + 1 );

		if ( a == NULL )

			return -3;

		*lineptr = a;

		*n = i;

	}

	/* read the data into the buffer */

	if ( fread( *lineptr, sizeof( char ), i, stream ) != i )

		return -4;

	( *lineptr )[i] = 0;

	/* return the number of chars read */

	return i;

}

/*

ssize_t getline (char **lineptr, size_t *n, FILE *stream)

This function reads an entire line from stream, storing the text (including the 

newline and a terminating null character) in a buffer and storing the buffer 

address in *lineptr.

Before calling getline, you should place in *lineptr the address of a buffer *n 

bytes long, allocated with malloc. If this buffer is long enough to hold the 

line, getline stores the line in this buffer. Otherwise, getline makes the 

buffer bigger using realloc, storing the new buffer address back in *lineptr and 

the increased size back in *n. See Unconstrained Allocation.

If you set *lineptr to a null pointer, and *n to zero, before the call, then 

getline allocates the initial buffer for you by calling malloc.

In either case, when getline returns, *lineptr is a char * which points to the 

text of the line.

When getline is successful, it returns the number of characters read (including 

the newline, but not including the terminating null). This value enables you to 

distinguish null characters that are part of the line from the null character 

inserted as a terminator.

This function is a GNU extension, but it is the recommended way to read lines 

from a stream. The alternative standard functions are unreliable.

If an error occurs or end of file is reached without any bytes read, getline 

returns -1. 

*/

ssize_t getline (char **lineptr, size_t *n, FILE *stream)

{

	return getdelim ( lineptr, n, '\n', stream );

}

Rewriting the Wheel: bin2hex – hex2bin

Yep everybody does it; repetition.  In the grand old days of old we all rebuilt the wheel leading to the establishment of patterns.  For me patterns often help me get to achieve my goals faster than using Google.  Using Google to find simple tools normally means understanding the search engine, browsing the results and evaluating each alternative in turn.  Invariably this means slightly altering someone else’s code, compromising objectives or installing a bulky tool that is bloated by a thousand functions you don’t need or want.  Naturally this last list is not exhaustive; just consider the dependencies that an external solution often requires.  If its simple and quick its often better to reinvent your wheel.

A perennial favuorite of mine is base translation; specifically hexadecimal to binary and binary to hexidecimal.  This is so simple and so useful it’s the subject of thousands of Google results.  Yet the useful bits are rarely on the first page of results.  So I coded it in about 30 minutes, and debugged it in about 30 minutes – because I was tired at the time.  Interestingly this is not the first time that I’ve written this code.  My earliest recollection of having written it was about 1986 over 22 years ago.  Of course it was coded in basic at the time on an Atari, however it wasn’t too long before I started translating it into C.

Patterns have long been a formal word in Computer Science.  In 1996 a very famous book, the name of which escapes me now and I’m too lazy to get it off my shelf, was dedicated to the development of patterns in software development.  Interestingly the way that the patterns were described was almost a deterent to the patterns themselves.  Being of academic nature the usefulness of the book was somewhat tarnished by its lack of pragmatic style and appeal to the great unwashed.  Its a fantastic book though which every software engineer should read.

For me patterns are best expressed in real world examples and in the languages that I understand.  It makes them instantly availble to me and much more likely to improve my work.  Yes they do need some formal definition most of the time.  However if its so simple as to be obvious then don’t clutter the example with an explanation other than to say what you used it for.  If its really that useful to be made into a pattern it’ll resurface again in the form of ‘now what did I do with that code I wrote 22 years ago that solved this problem?’

Well here it is, I used it to translate a Hexadecimal network packet lifted from an application log back into binary so that it could be replayed over a network for testing the application time and again.  Note the use of redirected standard input and output; its clean – no Swiss army knife in this one, what would I need a spoon on it for anyway?

hex2bin.c

#include "stdio.h"

#include "stdlib.h"

int main( int argc, char ** argv, char ** env ) {

  char h = '0';
  FILE * fi = stdin;
  FILE * fo = stdout;

  if ( argc > 1 ) {

    printf( "Usage: redirect input and output to stdin and stdout respectively.\n" );
    return 0;
  }

  h = getc( fi );

  while ( h != EOF ) {

    char b = 0;

    if ( h - '0' < 10 ) b = h - '0';
    else if ( h - 'a' < 'g' - 'a' ) b = ( h - 'a' ) + 10;
    else if ( h - 'A' < 'G' - 'A' ) b = ( h - 'A' ) + 10;

    b = b << 4;

    h = getc( fi );
    if ( h == EOF ) return 0;

    if ( h - '0' < 10 ) b += h - '0';
    else if ( h - 'a' < 'g' - 'a' ) b += ( h - 'a' ) + 10;
    else if ( h - 'A' < 'G' - 'A' ) b += ( h - 'A' ) + 10;

    if ( putc( b, fo ) == EOF ) return 0;

    h = getc( fi );
  }

  return 0;

}

Of course in this case it was just too tempting to write it’s sister as well: binary to hex. If you combine these two tools on the command line with netcat and some other favourite script favourites you can make quite a useful test tool.

bin2hex.c

#include "stdio.h"

#include "stdlib.h"

int main( int argc, char ** argv, char ** env ) {

  char b = '0';
  FILE * fi = stdin;
  FILE * fo = stdout;

  if ( argc > 1 ) {

    printf( "Usage: redirect input and output to stdin and stdout respectively.\n" );
    return 0;
  }

  b = getc( fi );

  while ( b != EOF ) {

    char h = 0;

    h = b >> 4;
    if ( h < 10 ) h = h + '0';
    else if ( h < 16 ) h = h + 'a';
    if ( putc( h, fo ) == EOF ) return 0;

    h = b & 0x0f;
    if ( h < 10 ) h = h + '0';
    else if ( h < 16 ) h = h + 'a';
    if ( putc( h, fo ) == EOF ) return 0;

    b = getc( fi );
  }

  return 0;

}

Lastly there is a point that is quite good to note about such simple patterns: they make really a really good basis for recruitment tests. You can always tell how good an organisation is in recruitment by the quality of this process. If they complain about spelling it generally means that the organisation is focussed on minutia and micro-management is probably the order of the day. Normally that’s a warning sign. So I always include spelling mistakes in my submissions. However if you are asked about the various ways of implementing the illustrated code (once of course you’ve rewritten it for them from requirements!) and integration strategies, impact on speed, memory usage along with testing and comparison of the requirements provided you’re probably to a good thing.

So just for fun let’s look at one aspect of this. IF you go do that dreaded Google search and sift through the cruft out there you will find a very specific approach often used that is quite different to that which I’ve used above to determine binary or hex value in translation; it normally involves using a switch statement like this:

char h;
int b;

switch ( h ) {
  '0'  :  b = 0;
          break;
  '1'  :  b = 1;
          break;
  '2'  :  b = 2;
          break;
  '3'  :  b = 3;
          break;
  '4'  :  b = 4;
          break;
  '5'  :  b = 5;
          break;
  '6'  :  b = 6;
          break;
  '7'  :  b = 7;
          break;
  '8'  :  b = 8;
          break;
  '9'  :  b = 9;
          break;
  'A'  :
  'a'  :  b = 10;
          break;
  'B'  :
  'b'  :  b = 11;
          break;
  'C'  :
  'c'  :  b = 12;
          break;
  'D'  :
  'd'  :  b = 13;
          break;
  'E'  :
  'e'  :  b = 14;
          break;
  'F'  :
  'f'  :  b = 15;
          break;
  default : return -1;
}

All this seems reasonable. Yes its going to be bigger than my code but its going to be quicker right? Actually that’s not right. You see the code is larger so its going to take a larger number of get requests to the memory. My code is small enough to fit inside most modern processor’s register sets not even touching the processor cache which the static example just given would have to sit in. Next the compiler on average will make twice the number of comparisons to jump into the switch statement, than the corresponding number of operations for my version. But this is also compiler and processor dependent. Coding simply in this case makes the optimiser’s job easier but an optimiser normally doesn’t have the nous to translate the code to another approach, just apply obvious short cuts to the code already presented. Naturally this could lead into discussions about optimisers and their abilities and limitations.

In fact there are a large number of optimisations that could be made to my code that would improve its performance too. But I’ll stop here because I could write a book on this one piece of code. So you see a particularly innocuous, been-done-before, simple piece of code can be interesting and be used to draw out the depth of a person’s knowledge. Plus there’s a certain amount of satisfaction in being able to go back an polish old code, it’s like visiting an old friend.


A Few Ajax Links

AJAX is a now a well established technology. It has been around for a long time but only recently (2005) was the term ‘AJAX’ coined to mean Asynchronous JavaScript And XML and unify the previously disjointed technologies it represents. Essentially it is an amalgamation of technologies to provide good user interface tools for Web browser applications.

Now of course there are a myriad of these toolkits, most born soon after the publishing to the AJAX article in 2005, and most not really suitable IMHO for real application. But they are all useful to look at. For me the most interesting ones are the following:

There is pretty good AJAX web site here if you need a place for starting to look for AJAX resources.


C++ Valarray Matrix Madness

Numeric STL container valarray began life partially as an attempt to make C++ appealing to the supercomputing community. At the time the big thing in those big machines was, the ironically named, vector processing. However the valarray fell by the wayside as the people driving its development left the STL development group. Perhaps they realised that it didn’t really fit 100% with the STL, or maybe they just got sidetracked; who knows. But it is still useful, and here are a few reasons why:

  • Can be used to write faster code for numeric spaces than possible with other STL types like the ubiquitous vector template.
  • Offers the coder the possibility of staying within the comfort zone of the STL and familiar object oriented concepts with a small speed sacrifice over hand carved C.
  • Allows library developers a way of writing optimized libraries for different environments so that coder can concentrate on the development at hand and not loose track in the complexity of the target environment.

So I’ve been playing around with valarray. It seemed that the best example to play with is that old classic the matrix. So here it is. Yes I know that there are some particularly hairy things wrong with it, but its not meant as a copy and paste solution. Its here as the results of a learning exercise and an example of what’s possible with valarray. There are one or two places which are not implemented or have not been tested, but you should be able to complete or test these by just looking at the rest of the examples. It should compile and run as is.

Have a look and have fun.

#include <fstream>
#include <iostream>
#include <iterator>
#include <string>
#include <valarray>
#include <vector>

/*------------------------------------------------------------------------------
  matrix2d interface

  This is the template for the matrix class.  Its not that fancy and there could
  be a lot more added to it but that was not the point of the exercise.  It is
  also limited by the number of types supported by the underlying valarray data
  container.  From an OO point of view its really not that good either given
  that a lot of the manipulation or generator methods really don't need to have
  access directly to the data part of the construct.  However on a practical
  basis including those methods in this class provide small benefits in speed
  and allow things to be a little more obvious.  There are also problems in the
  way that the generators return new matrix2d objects.  But I've no intention to
  fix them as this was just a learning exercise.  I've tossed the implementation
  of the methods into seperate compilation objects to make the interface cleaner
  for inspection purposes.  In keeping with the valarray perspective there is no
  bounds checking anywhere - you have been warned.
  ----------------------------------------------------------------------------*/

template<ypename T>

class matrix2d {
public:

  // creates based on the rows and data size
  matrix2d(std::size_t rows, std::valarray<T> data);
  // creates an empty rows x size matrix
  matrix2d(std::size_t rows, std::size_t cols);
  // direct initialisation - beware that rows x cols must equal data.size()
  matrix2d(std::size_t rows, std::size_t cols, std::valarray<T> data);

  // get the number of rows in the matrix2d
  std::size_t rows() const;
  // get the number of columns in the matrix2d
  std::size_t cols() const;
  // get a copy of the data in the matrix2d
  std::valarray<T> array() const;

  // retrieve the data from row r of the matrix
  std::valarray<T> row(std::size_t r) const;
  // retrieve the data from col c of the matrix
  std::valarray<T> col(std::size_t c) const;

  // retrieve refernce to the data from row r of the matrix
  std::slice_array<T> row(std::size_t r);
  // retrieve refernce to the data from col c of the matrix
  std::slice_array<T> col(std::size_t c);

  // basic item reference
  T & operator()(std::size_t r, std::size_t c);
  // basic item retrieval
  T operator()(std::size_t r, std::size_t c) const;

  // generate a matrix sort each row then sort the rows - UNIMPLEMENTED
  matrix2d<T> sort();
  // genetate a new matrix that is the transposition of this one
  matrix2d<T> transpose();
  // generate a new matrix with this matrix's data and sort each row
  matrix2d<T> rowItmSort();
  // generate a new matrix with this matrix's data and sort each row in reverse
  matrix2d<T> rowItmSortRev();
  // generate a new matrix with this matrix's data and sort each col
  matrix2d<T> colItmSort();
  // generate a new matrix with this matrix's data and sort each col in reverse
  matrix2d<T> colItmSortRev();

  // generate a new matrix of this one with m appended below
  matrix2d<T> appendRows(matrix2d<T> &m);
  // generate a new matrix of this one with m appended to the right
  matrix2d<T> appendCols(matrix2d<T> &m);
  // generate a matrix of this one, upper left corner at row t col l - UNTESTED
  matrix2d<T> extractMatrix2d(size_t t, size_t l, size_t w, size_t h);

protected:

  std::size_t rows_;
  std::size_t cols_;
  std::valarray<T> data_;

};

/*------------------------------------------------------------------------------
  matrix2d implementation
  ----------------------------------------------------------------------------*/

/*------------------------------------------------------------------------------
  matrix2d constructors
  ----------------------------------------------------------------------------*/

template<class T>
matrix2d<T>::matrix2d(std::size_t rows, std::valarray<T> data) :
rows_(rows), cols_(data.size() / rows), data_(data) {}

template<class T>
matrix2d<T>::matrix2d(std::size_t rows, std::size_t columns) :
rows_(rows), cols_(columns), data_(rows * columns) {}

template<class T>
matrix2d<T>::matrix2d(std::size_t rows, std::size_t columns,
std::valarray<T> data) :
rows_(rows), cols_(columns), data_(data) {}

/*------------------------------------------------------------------------------
  matrix2d operations
  ----------------------------------------------------------------------------*/

template<class T>
std::size_t matrix2d<T>::rows() const {
  return rows_;
}

template<class T>
std::size_t matrix2d<T>::cols() const {
  return cols_;
}

template<class T>
std::valarray<T> matrix2d<T>::array() const {
  return data_;
}

template<class T>
std::valarray<T> matrix2d<T>::row(std::size_t r) const {
  return data_[std::slice(r * cols(), cols(), 1)];
}

template<class T>
std::valarray<T> matrix2d<T>::col(std::size_t c) const {
  return data_[std::slice(c, rows(), cols())];
}

template<class T>
std::slice_array<T> matrix2d<T>::row(std::size_t r) {
  return data_[std::slice(r * cols(), cols(), 1)];
}

template<class T>
std::slice_array<T> matrix2d<T>::col(std::size_t c) {
  return data_[std::slice(c, rows(), cols())];
}

template<class T>
T& matrix2d<T>::operator()(std::size_t r, std::size_t c) {
  return data_[r * cols() + c];
}

template<class T>
T matrix2d<T>::operator()(std::size_t r, std::size_t c) const {
  return row(r)[c];
}

/*------------------------------------------------------------------------------
  matrix2d generators
  ----------------------------------------------------------------------------*/

template<class T>
matrix2d<T> matrix2d<T>::sort() {

  matrix2d<T> result(rows_, cols_);

  /* TO DO TO DO TO DO TO DO TO DO TO DO TO DO TO DO TO DO TO DO TO DO TO DO */

  return result;
}

template<class T>
matrix2d<T> matrix2d<T>::transpose() {

  matrix2d<T> result(cols_, rows_);

  for (std::size_t i = 0; i < rows_; ++i)
    result.col(i) = static_cast<std::valarray<T> > (row(i));

  return result;
}

template<class T>
matrix2d<T> matrix2d<T>::rowItmSort() {

  matrix2d<T> result(rows_, cols_);

  for (std::size_t i = 0; i < rows_; ++i) {

    std::valarray<T> x = static_cast<std::valarray<T> > (row(i));
    std::sort(&x[0], &x[cols_]);
    result.row(i) = x;
  }

  return result;
}

template<class T> bool rev (const T & a, const T & b) { return a > b; }

template<class T>
matrix2d<T> matrix2d<T>::rowItmSortRev() {

  matrix2d<T> result(rows_, cols_);

  for (std::size_t i = 0; i < rows_; ++i) {

    std::valarray<T> x = static_cast<std::valarray<T> > (row(i));
    std::sort(&x[0], &x[cols_], rev<T>);
    result.row(i) = x;
  }

  return result;
}

template<class T>
matrix2d<T> matrix2d<T>::colItmSort() {

  matrix2d<T> result(rows_, cols_);

  for (std::size_t i = 0; i < cols_; ++i) {

    std::valarray<T> x = static_cast<std::valarray<T> > (col(i));
    std::sort(&x[0], &x[rows_]);
    result.col(i) = x;
  }

  return result;
}

template<class T>
matrix2d<T> matrix2d<T>::colItmSortRev() {

  matrix2d<T> result(rows_, cols_);

  for (std::size_t i = 0; i < cols_; ++i) {

    std::valarray<T> x = static_cast<std::valarray<T> > (col(i));
    std::sort(&x[0], &x[rows_], rev<T>);
    result.col(i) = x;
  }

  return result;
}

template<class T>
matrix2d<T> matrix2d<T>::appendRows(matrix2d<T> &m) {

  matrix2d<T> result(rows_ + m.rows_, cols_);

  result.data_[std::slice(0, rows_ * cols_, 1)] = data_;
  result.data_[std::slice(rows_ * cols_, m.rows_ * m.cols_, 1)] = m.data_;

  return result;
}

template<class T>
matrix2d<T> matrix2d<T>::appendCols(matrix2d<T> &m) {

  matrix2d<T> result(rows_, cols_ + m.cols_);

  std::size_t s1[] = {rows_,cols_}; // shape of left matrix
  std::size_t p1[] = {result.cols_,1}; // position of left matrix in result
  std::size_t s2[] = {m.rows_,m.cols_}; // shape of right matrix
  std::size_t p2[] = {result.cols_,1}; // position or right matrix in result

  std::valarray<std::size_t> sv1(s1, 2);
  std::valarray<std::size_t> pv1(p1, 2);
  std::valarray<std::size_t> sv2(s2, 2);
  std::valarray<std::size_t> pv2(p2, 2);

  result.data_[std::gslice(0, sv1, pv1)] = data_; // copy left matrix into place
  result.data_[std::gslice(cols_, sv2, pv2)] = m.data_; // repeat for m

  return result;
}

template<class T>
matrix2d<T> matrix2d<T>::extractMatrix2d(size_t x, size_t y, size_t w,
size_t h) {

  /* TEST ME TEST ME TEST ME TEST ME TEST ME TEST ME TEST ME TEST ME TEST ME */

  matrix2d<T> result(h, w);

  size_t x2[] = {h, w}, s[] = {rows_, 1};
  std::valarray<size_t> xa(x2, 2), sa(s, 2);

  result.data_ = data_[(const std::gslice)std::gslice(y * rows_ + x, xa, sa)];

  return result;
}

/*------------------------------------------------------------------------------
  matrix2d general input output

  This is a simple class to help collect methods of serialising the matrix2d
  data forms.  Really they should be done another way but once again I don't
  really care as they are throw away code for the purpoese of demonstration
  only.
  ----------------------------------------------------------------------------*/

template<typename T>

class matrix2dio {
public:

  matrix2d<T> textToM2d(std::istream & is, size_t w);
  matrix2d<T> fileToM2d(std::string file, size_t w);
  void m2dToText(std::ostream & os, const matrix2d<T> & m);
  void printValarray(std::ostream & os, const std::valarray<int> & va);
};

/*------------------------------------------------------------------------------
  matrix2dio operations
  ----------------------------------------------------------------------------*/

typedef std::istream_iterator<int> int_istrm_iter;

template<class T>
matrix2d<T> matrix2dio<T>::textToM2d(std::istream & is, size_t w) {

  std::vector<T> v;
  std::valarray<T> a;

  if (!is.good() || is.eof())
    return;

  copy(int_istrm_iter(is), int_istrm_iter(), back_inserter(v));
  a.resize(v.size(), sizeof ( int));
  copy(v.begin(), v.end(), &a[0]);

  return new matrix2d<T > (w, a);
}

template<class T>
matrix2d<T> matrix2dio<T>::fileToM2d(std::string file, size_t w) {

  std::filebuf fb;
  std::istream is(&fb);

  fb.open(file.c_str(), std::ios::in);
  matrix2d<T> m = textToM2d(is, w);
  fb.close();

  return m;
}

template<class T>
void matrix2dio<T>::m2dToText(std::ostream & os, const matrix2d<T> & m) {

  size_t i = 0, j = 0;

  for (i = 0; i < m.rows(); i++) {

    std::valarray<T> r = m.row(i);
    os << r[0];

    for (j = 1; j < m.cols(); j++)
      std::cout << ' ' << r[j];

    os << '\n';
  }

  os << std::flush;
}

template<class T>
void matrix2dio<T>::printValarray(std::ostream & os,
const std::valarray<int> & va) {

  copy(&va[0], &va[va.size() - 1], std::ostream_iterator<T > (os, " "));
  os << va[va.size() - 1] << std::endl;
}

/*------------------------------------------------------------------------------
  matrix2d tests
  ----------------------------------------------------------------------------*/

void testConstructors() {

  std::cout << "\n\n\nRunning Constructor Tests\n\n";

  int a[] = { 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 };
  std::valarray<int> v( a, 12 );
  std::size_t x = 3, y = 4;
  matrix2dio<int> i;

  std::cout <<
    "\nTesting: matrix2d(std::size_t rows, std::size_t columns, "
    "std::valarray<T> data);\n";
  matrix2d<int> m1( x, y, v );

  std::cout << "number of rows: " << m1.rows() << '\n'
  << "number of cols: " << m1.cols() << '\n'
  << "matrix content:\n";
  i.m2dToText( std::cout, m1 );

  std::cout << "\nTesting: matrix2d(std::size_t rows, std::size_t columns);\n";
  matrix2d<int> m2( x, y );

  std::cout << "number of rows: " << m2.rows() << '\n'
  << "number of cols: " << m2.cols() << '\n'
  << "matrix content:\n";
  i.m2dToText( std::cout, m2 );

  std::cout <<
    "\nTesting: matrix2d(std::size_t rows, std::valarray<T> data);\n";
  matrix2d<int> m3( x, v );

  std::cout << "number of rows: " << m3.rows() << '\n'
  << "number of cols: " << m3.cols() << '\n'
  << "matrix content:\n";
  i.m2dToText( std::cout, m3 );
}

void testRowsAndCols() {

  std::cout << "\n\n\nRunning Row/Col Accessor Tests\n\n";

  int a[] = { 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 };
  std::valarray<int> v( a, 12 );
  std::size_t x = 3, y = 4;
  matrix2dio<int> i;

  matrix2d<int> m1( x, y, v );

  std::cout << "\nTesting: std::valarray<T> row(std::size_t r) const;\n";

  std::valarray<int> r1 = m1.row(0);
  std::valarray<int> r2 = m1.row(1);
  std::valarray<int> r3 = m1.row(2); 

  std::cout << "row 1:\n";
  i.printValarray( std::cout, r1 );
  std::cout << "row 2:\n";
  i.printValarray( std::cout, r2 );
  std::cout << "row 3:\n";
  i.printValarray( std::cout, r3 );

  std::cout << "\nTesting: std::valarray<T> col(std::size_t r) const;\n";

  std::valarray<int> c1 = m1.col(0);
  std::valarray<int> c2 = m1.col(1);
  std::valarray<int> c3 = m1.col(2);
  std::valarray<int> c4 = m1.col(3); 

  std::cout << "col 1:\n";
  i.printValarray( std::cout, c1 );
  std::cout << "col 2:\n";
  i.printValarray( std::cout, c2 );
  std::cout << "col 3:\n";
  i.printValarray( std::cout, c3 );
  std::cout << "col 4:\n";
  i.printValarray( std::cout, c4 );

  std::cout << "\nTesting: std::slice_array<T> row(std::size_t r);\n";

  std::slice_array<int> rs1 = m1.row(0);
  std::slice_array<int> rs2 = m1.row(1);
  std::slice_array<int> rs3 = m1.row(2); 

  std::cout << "row 1:\n";
  i.printValarray( std::cout, rs1 );
  std::cout << "row 2:\n";
  i.printValarray( std::cout, rs2 );
  std::cout << "row 3:\n";
  i.printValarray( std::cout, rs3 );

  std::cout << "\nTesting: std::slice_array<T> col(std::size_t r);\n";

  std::slice_array<int> cs1 = m1.col(0);
  std::slice_array<int> cs2 = m1.col(1);
  std::slice_array<int> cs3 = m1.col(2);
  std::slice_array<int> cs4 = m1.col(3); 

  std::cout << "col 1:\n";
  i.printValarray( std::cout, cs1 );
  std::cout << "col 2:\n";
  i.printValarray( std::cout, cs2 );
  std::cout << "col 3:\n";
  i.printValarray( std::cout, cs3 );
  std::cout << "col 4:\n";
  i.printValarray( std::cout, cs4 );
}

void testGenerators() {

  std::cout << "\n\n\nRunning Generator Tests\n\n";

  int a[] = { 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 };
  int b[] = { 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 110, 120 };
  std::valarray<int> v1( a, 12 );
  std::valarray<int> v2( b, 12 );
  std::size_t x = 3, y = 4;
  matrix2dio<int> i;

  matrix2d<int> m1( x, y, v1 );
  matrix2d<int> m2( x, y, v2 );

  std::cout << "\nTesting: matrix2d<T> transpose();\noriginal:\n";

  matrix2d<int> m3 = m1.transpose();
  i.m2dToText( std::cout, m1 );
  std::cout << "transposed:\n";
  i.m2dToText( std::cout, m3 );

  std::cout << "\nTesting: matrix2d<T> appendRows(matrix2d<T> &v);\n";

  matrix2d<int> m4 = m2.appendRows(m1);
  i.m2dToText( std::cout, m4 );

  std::cout << "\nTesting: matrix2d<T> appendCols(matrix2d<T> &m);\n";

  matrix2d<int> m5 = m2.appendCols(m1);
  i.m2dToText( std::cout, m5 );

  std::cout << "\nTesting: matrix2d<T> rowItmSort();\n";

  matrix2d<int> m6 = m5.rowItmSort();
  i.m2dToText( std::cout, m6 );

  std::cout << "\nTesting: matrix2d<T> rowItmSortRev();\n";

  matrix2d<int> m7 = m5.rowItmSortRev();
  i.m2dToText( std::cout, m7 );

  std::cout << "\nTesting: matrix2d<T> colItmSort();\n";

  matrix2d<int> m8 = m4.colItmSort();
  i.m2dToText( std::cout, m8 );

  std::cout << "\nTesting: matrix2d<T> colItmSortRev();\n";

  matrix2d<int> m9 = m4.colItmSortRev();
  i.m2dToText( std::cout, m9 );
}

void testMatrix2d() {

    testConstructors();
    testRowsAndCols();
    testGenerators();
}

/*------------------------------------------------------------------------------
  matrix2d test entry point
  ----------------------------------------------------------------------------*/

int main( int argc, char** argv ) {

  testMatrix2d();
  return (EXIT_SUCCESS);
}

Fussing with slice_array

Some times we just make things far to complicated for ourselves: note to self (and whoever cares),do not put compiler hints into unproven code. Every now and then you overstep your confidence and do something that you look back at and think, “Now that’s just stupid!”. I’ve just spent the last several days wondering why the following does not compile.

void foo( const std::valarray<int> & va ) {
  std::slice_array<int> sa = va[std::slice( 1, 1, 1 )];
}

After all the line in the middle is exactly the same as sooo much example code out there it isn’t funny. However the compiler throws an awful wobbly and kicks back an error like this.

example.cc: In function ‘void foo(const std::valarray<int>&)’:
example.cc:20: error:
    conversion from
        ‘std::_Expr<std::_SClos<std::_ValArray, int>, int>’
    to non-scalar type
        ‘std::slice_array<int>’
    requested

Which is not quite what I expected. I’ve laid it out a bit different to the compiler for readability. Essentially the error means that the type of the expression on the right is not equivalent to the type of the variable on the left. However you can do this.

void foo( const std::valarray<int> & va ) {
  std::valarray<int> sa = va[std::slice( 1, 1, 1 )];
}

Which of course will compile without issue. OK so the operator() is overloaded, what’s its problem? The solution is blindingly obvious, but for those with a sense of anticipation I’ll drag it out just a tiny wee bit more. The mystery, and partly where I went wrong, was deepened by the valarray header.

/**
*  @brief  Return an array subset.
*
*  Returns a new valarray containing the elements of the array
*  indicated by the slice argument.  The new valarray has the same size
*  as the input slice.  @see slice.
*
*  @param  s  The source slice.
*  @return  New valarray containing elements in @a s.
*/
_Expr<_SClos<_ValArray, _Tp>, _Tp> operator[](slice) const;

/**
* @brief Return a reference to an array subset.
*
* Returns a new valarray containing the elements of the array
* indicated by the slice argument. The new valarray has the same size
* as the input slice. @see slice.
*
* @param s The source slice.
* @return New valarray containing elements in @a s.
*/
slice_array<_Tp> operator[](slice);

Now that's probably give it away if you didn't already know. But once again I wasn't looking properly. There were two real problems and none of them to do with my code: firstly I was programming by Google, secondly I was trying to be smart. The first error was as stupid as a cut and paste error. I'd looked up similar issues in Google and cut someone else's understanding of the problem out of a page that I had read and stapled it into my brain. Seems someone else had the same problem but not the nouse to get over it and was blaming the spec. Very silly, I should have known better. The second thing was that I'd tried to hint to the compiler before understanding its implications. Never hint to the compiler before you have the thing working, I'd have never wasted my time with it, although programming by debugger has its own worms.

Turns out this is the code that I wanted.

void foo( std::valarray<int> & va ) {
  std::slice_array<int> sa = va[std::slice( 1, 1, 1 )];
}

See it? If you still can't see it compare it to the earlier version and have a look at the prototypes in copied from the header above, which should explain almost everything. All that time. Silly, silly silly ... proceed with self flagellation of 50 lashings with a wet noodle! I have 15 years experience programming with C++ and 20 with C (which can generate similar errors), very embarrassing.


Shrinking by GIMP Lisp

Image alteration really annoys me some times when the tools that are on offer to do a cheap job show their price through the results. In other words there have been times that my efforts to shrink and enlarge images have resulted in a, rough, artifact chocked image of a quality annoying to the eye. Normally this is the result of some cheap Python script which has worked well enough in the past for generating eyecons but doesn’t really scale that well. Naturally there are better tools out there like the GIMP that I can alter images with but they come at a price: manual process = my time. This can be very time consuming when you have a whole directory full of images of my children that need to be processed from just one weekend. In walks the GIMP scritps to save the day.

GIMP scripts really really appeal to me for two reasons; firstly I love the GIMP as a tool (ignoring its low image bit rate), but more importantly the scripting language of choice is Lisp. Lisp is cool. Don’t ask me why I really don’t know, I just really like it and have ever since I first met it on a DEC Vax at university about fifteen years ago. I’d almost assigned it to the place where Prolog has gone to die when I realised that its at the back of my favourite text editor Emacs. However Emacs has most of its useful stuff already written so I’ve not really had the cause to play in Lisp till now.

So last year I dug into the GIMP and Lisp and came up with a couple of scripts that I like to keep in ‘.gimp-2.2/scripts’ in my home directory. They run over a list of images or a single and perform a nice image shrink. We don’t need to to a scale up as our camera takes images that are too large to be useful on a screen anyway. So here’s the script.

image-shrink.scm

;file: image-shrink.scm
;location: ~/.gimp-2.2/scripts
;version: 0.1
;date: 2007-06-05
;
;definition:  single image shrink
;explanation: shrinks proportionally by width
;examples:
;  rm bike.jpg && cp bike.34.jpg bike.jpg && gimp -i -b '(image-shrink "bike.jpg" 500)' '(gimp-quit 0)'
;  gimp -i -b '(image-shrink "/home/michael/Desktop/bike.jpg" 500)' '(gimp-quit 0)'
;code:
(define
  (image-shrink filename width)
  (let*	(
          (image    (car (gimp-file-load RUN-NONINTERACTIVE filename filename)))
          (drawable (car (gimp-image-get-active-layer image)))
          (iheight  (car (gimp-image-height image)))
          (iwidth   (car (gimp-image-width image)))
        )
        (gimp-drawable-transform-scale
           drawable
           0.0 0.0
           width
;          iheight
           (* iheight (/ width iwidth))
           0 2
           1 3 0
	)
        (gimp-file-save RUN-NONINTERACTIVE image drawable filename filename)
        (gimp-image-delete image)
    )
)
;
;definition:  multi-image shrink
;explanation: calls image-shrink repeatedly to shrink files from a globbed list
;examples:
;  rm bike.jpg && cp images/bike.full.34.jpg bike.jpg && gimp -i -b '(images-shrink "/home/michael/Desktop/bike.jpg" 500)' '(gimp-quit 0)'
;  gimp -i -b '(images-shrink "/home/michael/Desktop/bike.jpg" 500)' '(gimp-quit 0)'
;code:
(define
  (images-shrink pathname width)
  (let* (
          (filelist (cadr (file-glob path 1)))
        )
        (while
          (not (null? filelist))
          (let* (
                  (filename (car filelist))
                )
                (image-shrink filename width)
          )
          (set! filelist (cdr filelist))
        )
  )
)

Nautilus Image Scripts

My wife and I do a lot of image processing of our family photos plus the photos of Lisa’s balloons. The most common things that we want to do is to rotate images by 90 degrees to make a portrait photo upright and rename the imagest to be named by the date and name. To do this I simply wrote a few bash scripts and put them in the ‘.gnome2/nautilus-scripts’ directory, then they appeared in the scripts list of the right click context menu in the nautilus file browser windows. This is especially useful as it allows you to select a whole bunch of photos and right click them to get them changed all at once. Note that the scripts do presume a hard wired temporary directory to use for the transforms and also require the exiftool to be accessible on the system.

name_to_date_and_time.sh

#!/bin/bash
TMP_FILE=`tempfile 2> /dev/null` || TMP_FILE="/tmp/nautilus-script.$$"
IFS="
"

trap "rm -f $TMP_FILE" EXIT

for F in $NAUTILUS_SCRIPT_SELECTED_FILE_PATHS; do
  cd `dirname $F`
  mv $F `exiftool -T -d "%Y-%m-%d %H-%M-%S" -createdate $F`
done

rotate_left.sh

#!/bin/bash
TMP_FILE=`tempfile 2> /dev/null` || TMP_FILE="/tmp/nautilus-script.$$"
IFS="
"

trap "rm -f $TMP_FILE" EXIT

for F in $NAUTILUS_SCRIPT_SELECTED_FILE_PATHS; do
  cd `dirname $F`
  mv -f $F $TMP_FILE
  jpegtran -copy all -rotate 270 -outfile "$F" $TMP_FILE
done

rotate_right.sh

#!/bin/bash
TMP_FILE=`tempfile 2> /dev/null` || TMP_FILE="/tmp/nautilus-script.$$"
IFS="
"

trap "rm -f $TMP_FILE" EXIT

for F in $NAUTILUS_SCRIPT_SELECTED_FILE_PATHS; do
  cd `dirname $F`
  mv -f $F $TMP_FILE
  jpegtran -copy all -rotate 90  -outfile "$F" $TMP_FILE
done

Generating Combinations

One of the tasks of numerical analysis is to search for or compare patterns. In the deep past this was quite a problem as it was easy to swamp the systems available with data making it more practical to employ a mathematician to perform some arcane analysis via algebra. However the manual way did not give the code opportunity to be tested against all possible combinations. Normally tests were conductd against boundaries and samples of statisitically significant combinations.

However in these days of Gig’s of memory and fast processors not to mention the terrabytes of disk space available cheaply it is possible to generate and test every possible combination in many cases. One of these is the classic combination space of mathematics. This following alogorithm is a crude base to progress from to generate such a number space.

// pattern generation
void genPat(size_t smpSize, size_t popSize, vector cmbSpace) {
  t_combo v(smpSize); // entry vector
  t_combo m(smpSize); // entry limit vector

  int i, e, a; // index and place holders

  // initialise the entry vector with an ascending order 1 to smpSize
  // and the limit vector with the max value each entry should reach
  for (i = 0; i < smpSize; i++)
  {
    v[i] = i + 1;
    m[i] = popSize - smpSize + v[i];
  }

  // repeat until the first entry in the vector is less than its
  // maximum value
  while (v[0] < m[0])
  {
    // update the combination space
    cmbSpace.push_back(v);

    // set an index to the last element
    e = smpSize - 1;

    // find the first element from the right to be less than the limit
    while (v[e] == m[e]) e--;

    // add one to and store the value of the current element in a var
    a = ++v[e];

    // set each element to the right to one more than the current
    while (e < smpSize - 1) v[++e] = ++a;
  }

  cmbSpace.push_back(v);
}

Copyright © 1996-2010 Code Snips. All rights reserved.
iDream theme by Templates Next | Powered by WordPress