Saturday, 15 August 2015

Preventing SQL Injection without using bind variables

SQL Injection can be a serious threat to your database. Depending of the setup of your database an attacker could destroy your data or get hold of sensitive data.
Many posts have been written about SQL Injection in Apex and the solution always has been to use bind variables instead of using substitution. There can be cases in which using bind variables is not an option or requires a lot more effort.

One example of this is the use of apex_util.json_from_sql :
declare 
  v_sql varchar2(4000);
begin
    v_sql := q'!select * from emp where ename like '#EMP#' !';
    v_sql := replace(v_sql,'#EMP#',apex_application.x01);
    apex_util.json_from_sql (v_sql);
end;
In the code above you can see that we need to incorporate the parameter into the query. There is no way we can use bind variables. This puts the door wide open to SQL Injection attacks.
However, if we split the parameter into an expression of which the result is equal to the parameter, no harm can be done. One possibility is to take the separate characters of the parameter and concatenate them.
parameter  := 'JON';
expression := 'J' ||'O'||'N';
The expression can be substituted in the query instead of the litteral parameter and you query is safe for SQL Injection.

The following procedure does the trick for you:
  function split_param ( p_param in varchar2) return varchar2 is
    l_return    varchar2(4000) := null;
    quote       char(1)        := chr(39);
  begin
    for i in 1..length(p_param) loop
      if l_return is not null then
        l_return := l_return || '||';
      end if;
      l_return := l_return || quote || substr(p_param,i,1) || quote;
    end loop;
    l_return := ltrim(rtrim(l_return,quote),quote);
    return(l_return);
  end;
This function accepts a string and outputs an expression that returns the same string.
We can use this function to convert the input parameter into an expression. This way the code will not be vulnerable to SQL Injection any more:

declare 
  v_sql varchar2(4000);
begin
    v_sql := q'!select * from emp where ename like '#EMP#' !';
    v_sql := replace(v_sql,'#EMP#',split_param(apex_application.x01));
    apex_util.json_from_sql (v_sql);
end;
I think it cannot be broken anymore. Do you agree or is this code still vulnerable to SQL Injection?

Love to hear from you!

Happy Apexing

Postscript: I have got a reaction on this blogpost by Carsten Czarski that notes that the dbms_assert package also contains functions to prevent SQL Injection. The function dbms_assert.enquote_literal ensures that all quotes within a string are paired, otherwise an exception is raised.
Thank you, Carsten.

Monday, 22 June 2015

Single Page Application with Oracle Apex

This post is the summary of a presentation held at KScope 2015. You will find links at the bottom to download the presentation and the example application containing all the code.

When using a web application on a smartphone a slow network can be really annoying. Especially moving to another web page can take more than a few seconds.
In mobile development often Single Page Applications are the answer to speed up the application. Because the whole application is on one page the only requests to the server are for getting data or submitting transactions.

Because my Apex smartphone application performed badly I wanted to convert it to a Single Page Application. As this is not supported by Apex the actions needed had to be coded manually in PL/SQL and JavaScript.

The Apex architecture relies strongly on page refreshes. Most of the processing is done when rendering the page or when submitting a page. All this processing can not be used in a SPA because it is rendered only once and never submitted (the Apex way). The image below shows which part of the Apex functionality is available.




There Dynamic Actions come to the rescue. They enable us to execute PL/SQL from an Apex page without having to code all the plumbing underneath ( JavaScript / Ajax / Serverside actions ).

So the way to go is to:
- create one page with all the regions we want on it
- disable all fetch processes
- replace page links with Javascript calls
- implement transactions with Dynamic Actions
- using JavaScript to handle (error)messages

The sample application contains some generic JavaScript functions for region management and messaging. On top of these functions for each region a set of specific JavaScript functions is built.

Much if not all of the code written could be generated from the Apex repository. It is possible to replace the manually coded PL/SQL performing the transactions by generic code based on the Apex repository. Also validations could be combined with this code. All the page links can be replaced by JavaScript calls. The messaging framework is already provided.

I think this way it is possible to generically create Single Page Applications with Apex, without all the manual coding described in the presentation. Only, I have not got the time to code it. Besides, it would be better if it were built into Apex, wouldn't it? Then everybody could benefit. Well Apex team, what about another challenge?

More information about the exact implementation is available in the presentation : SPA with Oracle Apex You can also download the example application ( Apex 5, only the EMP table needed ) : http://www.detora.nl/downloads/kscope2015/apex_spa.sql
Be sure to check the supported objects when importing because the application depends on a package spa_pck.sql. 

Happy Apexing,
Dick Dral






Thursday, 12 February 2015

Time input by touch movements


Since the early beginnings of the smartphone era I have been thinking about time input by drawing the hands of a clock on the screen of a smartphone. For me this works a lot easier than entering the numbers on the on-screen keyboard. As for intuitive usage, most of us have learned to read the clock in school, so it is very familiar. The only difference is that you 'write' the clock instead of reading it. 

About a year ago I had the time and skills to create a prototype. It functioned but lacked the ‘looks’ and was not very intuitive to use. Lately I had some inspiration to  restyle the original design to make it prettier and more intuitive. I would like to proudly present the result to you:



The image above will not respond to your touch, but the next link will take you to a page where you can try this control.

http://www.detora.nl/tic_demo/tic_demo.html

The control is designed for touch devices. In a desktop browser you can click on the numbers in the analog clock to select the hours or minutes.
If you want to try the control on your phone or tablet you can have an email with a link to the demo page sent to you. Enter your email address below and press the 'Send link by mail'. Your mail client will open and you can send the mail.

Send to email address     Send link by mail
( Your email address is not stored )

Below is an explanation how to use the control:
Touch the clock image next to the time input field and the time input control will appear.
In the center is the analog clock in which the hands can be drawn. Above it you see the digital time with buttons for changing to AM or PM. Below it are buttons to Accept of Cancel. The active hand and the active part of the digital display are indicated by a red color.

Draw the hands starting (approximately) at the center of the clock. To enter 10:15 you start at the center of the clock and swipe towards the 10, then release. The hour in the digital time display above the clock will display 10 and the minutes will have turned red (=active). Then start in the center of the clock again moving toward the 3 this time. When you release you will notice that the digital clock displays 10:15. To change to PM touch the PM button.
If you are satisfied you can press OK to save the time. Cancel returns to the form without saving the time.

I would really like to hear whether you think this control useful. Or maybe you have  suggestions on improving it.

Happy developing,


Thursday, 8 January 2015

Using a pipelined function with parameters in a view

In the Oracle database you can use pipelined functions to improve performance or decrease complexity. With the table function a pipelined functions can be called:

select * from table(pf_demo(10));
This way it is not possible to wrap the parameterised pipelined function in a view. You can only create a view for a specific parameter value:

create view pf_demo_view  as select * from table(pf_demo(10));
This is not very useful. Luckily there is a way to pass parameters to a pipelined function within a view. This solution uses package variables to store the parameter values and setter functions to set the values.
In our example we create the pipelined function in a package pf_demo_pkg. In the case of NULL-arguments the values of package variables are used. We create the view with a call to the pipelined function without parameters. The view can be queried with a call to the setter function(s) in the where clause. In the case of NULL-arguments the values of package variables are used. We create the view with a call to the pipelined function without parameters. The view can be queried with a call to the setter function(s) in the where clause. This way you can query the view from PL/SQL.
create view pf_demo_view  as select * from table(pf_demo());

begin
  pf_demo_pkg.set_dept(10);
  for r in ( select * from pf_demo_view ) loop
    ...
  end loop;
end;
Now there is also a solution to use such a view from SQL. This solution utilizes the fact that the where clause of a query is interpreted before the from clause. So when we call the setter(s) in the where-clause the view will pass these parameters to the piplined function. You can see this in action in the code below:

SQL 
SQL create  type emp_rowtype
  2   as object
  3   ( empno number
  4   , ename varchar2(10)
  5   , hiredate date
  6   , deptno number
  7  );
  8  /

Type created.

SQL 
SQL create  type emp_table_type as table of emp_rowtype;
  2  /

Type created.

SQL 
SQL 
SQL create or replace package pf_demo is
  2  
  3    function set_deptno ( p_deptno in number) return number;
  4  
  5    function emp_pf (p_deptno in number default null)return emp_table_type  pipelined;
  6  
  7  end;
  8  /

Package created.

SQL 
SQL  create or replace
  2  package body pf_demo is
  3  
  4    g_deptno    number := null;
  5  
  6    function set_deptno ( p_deptno in number) return number is
  7    begin
  8    g_deptno := p_deptno;
  9         return(p_deptno);
 10    end;
 11  
 12    function emp_pf (p_deptno in number default null)return emp_table_type  pipelined is
 13    cursor c_emp (cp_deptno number) is
 14    select empno
 15         , ename
 16         , hiredate
 17         , deptno
 18    from emp
 19    where deptno = cp_deptno
 20   ;
 21   r_emp     c_emp%rowtype;
 22   r_rec     emp_rowtype;
 23   l_empno     number := null;
 24   l_ename     varchar2(10) := null;
 25   l_hiredate  date := null;
 26   l_deptno    number := null;
 27  begin
 28   l_deptno := nvl(p_deptno,g_deptno);
 29   open c_emp(l_deptno);
 30   loop
 31     fetch c_emp into  l_empno
 32          ,  l_ename
 33          ,  l_hiredate
 34          ,  l_deptno
 35     ;
 36     exit when c_emp%notfound;
 37     r_rec := new emp_rowtype
 38           ( l_empno
 39           , l_ename
 40           , l_hiredate
 41           , l_deptno
 42           );
 43     pipe row (r_rec);
 44   end loop;
 45   close c_emp;
 46   return;
 47   end;
 48  
 49  end pf_demo;
 50  /

Package body created.

SQL 
SQL 
SQL select * from table(pf_demo.emp_pf(10));

     EMPNO ENAME      HIREDATE      DEPTNO                                      
---------- ---------- --------- ----------                                      
      7782 CLARK      09-JUN-81         10                                      
      7839 KING       17-NOV-81         10                                      
      7934 MILLER     23-JAN-82         10                                      

SQL 
SQL select * from table(pf_demo.emp_pf()) where pf_demo.set_deptno(20) is not null;

     EMPNO ENAME      HIREDATE      DEPTNO                                      
---------- ---------- --------- ----------                                      
      7369 SMITH      17-DEC-80         20                                      
      7566 JONES      02-APR-81         20                                      
      7788 SCOTT      09-DEC-82         20                                      
      7876 ADAMS      12-JAN-83         20                                      
      7902 FORD       03-DEC-81         20                                      

SQL 
SQL create view pf_demo_view as select * from table(pf_demo.emp_pf());

View created.

SQL 
SQL select * from pf_demo_view where pf_demo.set_deptno(30) is not null;

     EMPNO ENAME      HIREDATE      DEPTNO                                      
---------- ---------- --------- ----------                                      
      7499 ALLEN      20-FEB-81         30                                      
      7521 WARD2      22-FEB-81         30                                      
      7654 MARTIN     28-SEP-81         30                                      
      7698 BLAKE      01-MAY-81         30                                      
      7844 TURNER     08-SEP-81         30                                      
      7900 JAMES      03-DEC-81         30                                      

6 rows selected.

SQL 
SQL spool off
Hope you can use this, happy coding Dick

Friday, 18 July 2014

Avoiding ORA-04068

This is a nasty error that can occur when a package is recompiled during an Oracle session. I work now in an environment where this occasionally happens and sometimes minutes of data entry are lost.
The same statement will execute without problem when re-issued but for this the software needs to be  thoroughly revised. 
The ORA-04068 problem is explained very well in a number of posts like Avoiding ORA-04068: existing state of packages has been discarded. One of the solutions is avoiding state fullness of packages, i.e. no global constants or variables. When inspecting the packages in the system I found that quite a number of packages only had global constants and no variables. At first I thought about moving the constants to a separate package. Then only changes to this package would cause a ORA-04068. The drawback of this solution is that code throughout the packages would have to be modified. 
Another solution is to replace the constants by functions. This way there is no global state and no code to be changed when the functions get the same name as the constants. This also works for public constants. So
 
 LF      VARCHAR2(10) := CHR(10);

is changed to
 
  FUNCTION LF RETURNS VARCHAR2 IS
  BEGIN
    RETURN(CHR(10));
  END;

To make things even easier I have coded a conversion utility where you can input your constant declarations a retrieve the corresponding functions. You find it here.

Enjoy!





Monday, 24 March 2014

The advantage of fonts over images

-->
Imagine your boss wants you to create an application with iconic buttons, He has an image that is to be used for menu buttons, and for the landing page, and as an image on some pages.



You start editing the image with photoshop creating the neccesary images. After some time you are ready. Some days later your boss says, there has bene a feedback session with the user and they want the buttons changed, different color with gradient effect and rouneded corners, and they should be a bit larger:


Again you start up PhotoShop and change the images. And you are lucky if this is the only time the specifications change.

All this image editing could be avoided when you would use images from a font. In a previous post I described how to create iconic buttons using existing fonts.  But it is also possible to define your own custom font containing the images you need ( how do you think Font Awesome is made?). 
Above appearances are all images generated using CSS and only one character in a custom font ( unfortunately I am not able to load the font in this blog).

The images you see in this blog are font characters styled with CSS. It is very easy to change the color and size of the image or the background, just change the CSS properties.

When you have a single color image you can incorporate it in a font. You need the image in SVG format, You can use tools like Inkscape to convert your image to SVG.
With your images in SVG you can build a font from them using icomoon.

It takes some time and effort to creatie a custom font. I did not have the time nor the patience to do it so I asked my 17-year old son Michiel to do it for me. He is  very good at this kind of thing (amongst others) and he has introduced me to many interesting new developments, like AngularJS and NodeJS. If you can use his help you can reach him here.





Tuesday, 18 February 2014

Splitting an Apex report into several columns

The other day I was asked to create a page where a large number of properties should be connected to an object. I created a report consisting of the rows with the property name and a check box. This resulted in a very long and narrow report. All of the data could not be seen without scrolling. So I decided the report should be split into a number of columns. And I wanted to avoid special constructs in Apex and SQL (I have done that before and it was very compex). So I decided to use JavaScript to move the rows. By using JavaScript the Apex transaction mechanism keeps functioning, Apex does not know or notice that anything in the layout of the page has been changed.



I want to achieve the right side situation starting with the left report ( that by the way continues a long way below where the picture ends)

With JavaScript a number of extra table elements are created on the same level as the table element containing the report. The thead element of the original table element will be copied to the other table nodes to provide the same heading and a tbody element will be created to house the rows. After that the rows can be distributed among the tables. With CSS the table elements are positioned next to each other. The only limit to the number of columns is the available horizontal space.
You should call the JavaScript function below at startup or in the refresh event when using partial page rendering.

function reportToColumns ( tabSelector, numCols )
{ var numRows = 0;
  // bereken aantal rijen per kolom
  var rowsPerCol = Math.ceil( ( $(tabSelector+' tr').length-1 )/numCols);
  var baseName = 'reportColumn'; 

  $(tabSelector).addClass('reportColumn');
  // maak kolommen aan
  for ( i = 2; i <= numCols; i++)
  { $(tabSelector).parent().append('
'); $(tabSelector+' thead').clone().appendTo( $('#'+baseName+i) ); $('#'+baseName+i).append(''); } // verdeel de rijen over de kolommen var id = 2; var dest = ''; $(tabSelector+' tr').each( function(index) { if (index > rowsPerCol) { dest = '#'+baseName+id+' tbody'; $(dest).append( $(this) ); numRows = numRows + 1; // wissel de kolom als rijen per kolom is bereikt if ( numRows >= rowsPerCol) { id = id + 1; numRows = 0; } } } ); }

I used classes for the layout of a specific report. You might need to change that according to your needs.
To make sure the div’s are placed next to each other a little bit of CSS is needed:
.reportColumn {
    display: block;
    float: left;
    margin-right: 30px;
    position: relative;
}

You can see this in action right here.
Note 03-10-2015: The software is dependent of the template used. The link above is suited for Theme 25. You will find a version for the Universal Theme here.

This method can be used on either Apex Reports or Apex Tabular Forms. In fact it is not a specific Apex solution and can be used on any HTML table. Happy Apexing