Properly handle VK_DELETE when subclassing edit control

Question

0.00/5 (No votes)

See more:

I am subclassing edit control to accept only decimal numbers.

Everything works fine, and the only thing left for me is processing the case when user selects portion of the text and presses Delete key ( VK_DELETE ).

My question is following:

Does pressing a Delete key ( VK_DELETE ) sends some message like WM_CLEAR that I can use for my processing, or do I have to handle WM_KEYDOWN in my subclassing procedure and do the processing there?

Thank you.

Best regards.

Posted 15-Mar-14 20:40pm

AlwaysLearningNewStuff

Add a Solution

Comments

Sergey Alexandrovich Kryukov 16-Mar-14 13:10pm

You can easily find it out using message spy application. Not clear why would you need to handle VK_DELETE in any non-default way at all. You only need to allow decimal digits as input characters (and, don't forget, backspace, the "character" 0x08).
—SA

AlwaysLearningNewStuff 16-Mar-14 22:47pm

Mr.Kryukov,

First off thank you for taking time to see this post and thank you for replying.

This will take a while so please forgive me for a lengthy comment. If you have further questions I will understand, and I will answer them as always.

Here is the problem:

I have a dialog box with lots of edit controls, radio buttons and other stuff and it has a purpose to communicate with a database.

Dialog box can both read and edit the data from a database, and there is the problem.

The maximal allowed length of the decimal number is 10, and here is the critical case:

Let us say user enters .123456789 into database. This will be stored as 0.123456789 which is longer than maximal allowed length by 1 character ( the leading zero was added ).

User may make a mistake by clicking the wrong radio button, for example, and will decide to correct it. My dialog box has an option of reloading the data and user will do that in order to make the above mentioned correction.

Now I have a dilemma:

If I truncate the data I will lose the part of the information that may be important. I consider this a bad coding and unacceptable option.

If I just load the data into edit control it will have 0.123456789 text which is 11 characters long.

Imagine that user corrects the before mentioned mistake by clicking on the correct radio button and hits "Edit" button.

Now I will use GetWindowText( hwndOfMyEditControl, myStringVariable, 10 ); to store back the text from the edit control and get buffer overflow since eleven characters can't stay in an array of 10 characters.

Therefore to fix this problem I have decided to forbid the user to leave out the leading numbers.

This makes -.23 and other similar formats invalid.

If user enters -123.456 ( where red text represents current selection ) and presses Delete, I will get -.23. The user may continue to add numbers at the end of the number, thus getting the critical case I have described ( he could just add seven consecutive nines and get something like -.239999999 which is eleven characters long and again I will get buffer overflow ).

I do not how to handle this case properly, so I have decided to subclass the edit control and just forbid the format where leading numbers are left out, as I have already mentioned.

That is why I needed to properly handle every deleting ( WM_CUT, WM_CLEAR, backspace and Delete ).

I have posted this question on StackOverflow as well, and got an interesting answer:

The poster used Spy++ on Notepad to actually determine that only WM_KEYDOWN and WM_KEYUP messages are sent when user presses Delete. Therefore I have implemented my handling of Delete via WM_KEYDOWN.

If you have a better way to reccomend for protection of the above mentioned buffer overflow, I would highly appreciate it, as always. After all we learn new and useful things aslong as we live, right?

I aain apologize for the lengthy comment, but I wanted to explain the problem properly, and I never spare "words" to do that.

Best regards.

1 solution

Add a Solution

Add your solution here

Treat my content as plain text, not as HTML

Preview 0

…

Existing Members

Sign in to your account

...or Join us

Download, Vote, Comment, Publish.

Your Email
Password
Forgot your password?

Your Email
This email is in use. Do you need your password?
Optional Password

I have read and agree to the Terms of Service and Privacy Policy
Please subscribe me to the CodeProject newsletters

When answering a question please:

Read the question carefully.
Understand that English isn't everyone's first language so be lenient of bad spelling and grammar.
If a question is poorly phrased then either ask for clarification, ignore it, or edit the question and fix the problem. Insults are not welcome.
Don't tell someone to read the manual. Chances are they have and don't get it. Provide an answer or move on to the next question.

Let's work to help developers, not make them feel stupid.

This content, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)

Sergey Alexandrovich Kryukov · Accepted Answer · 2014-03-16T17:24:00

Solution 1

The detail explained in your comment are more important than the question itself, where you did not tell us the most important things. The problem is purely artificial, and here is the root of it: "This will be stored as 0.123456789 which is longer than maximal allowed". That said, you are string strings representing numbers in the database. This is totally wrong, wasteful, unreliable. You need to store data of numeric types. Generally, this is a weird trend haunting many beginners these days: working with strings representing data instead of data itself. This is very bad.

And of course, in the UI, you should better filter out unwanted characters. For example, if you enter fractional data, you may need to allow only digits, decimal delimiter (remember, it depends on culture!) and, optionally + or - (you may want to make it depending on the allowed range), and also backspace. This filtering is the routine techniques. I don't think this is a problem for you.

As to handling VK_DELETE, you don't need to do anything but default. Please see my comment to the question.

—SA

Posted 16-Mar-14 17:24pm

Sergey Alexandrovich Kryukov

Comments

AlwaysLearningNewStuff 17-Mar-14 1:13am

There is a misunderstanding again between us. Let us try to clarify matters again:

I use _wtof_l( myDoubleVariable, _get_current_locale() ) to properly convert string to double.

I insert into database using ADO ( I havent learned ADO.NET yet ), by using Command object.

Everything works fine. When I type in edit control .23 my query successfully executes.

The point is that once I open database to check if it really executed well, I see 0.23 which is OK, since wtof adds automatically the leading zero.

If I now load this value into edit control I get 0.23 text, which is also OK, since that is a number stored in database.

However if I call GetWindowText there is a danger of buffer oferflow, due to reasons described in my comment.

Hopefully this clarifies thing better.

Thank you Mr.Kryukov for trying to help and for your patience.

Best regards.

Sergey Alexandrovich Kryukov 17-Mar-14 10:03am

Why do you have strings in first place? To convert it to double? Don't store strings.
—SA

AlwaysLearningNewStuff 17-Mar-14 17:07pm

I have a database with different fields like employee names, consumption of electric energy, phone number, company name etc.

I store numeric data as numbers, of course, but GetWindowText stores the text of the edit control into a char array ( please see the documentation ) and that is the only way to get the text from the edit control if it is a decimal number.

Since edit control's text must be stored in a string I must convert it to double, so I can properly store it in a database.

Again, let me clarify:

The database field is of type Double, INSERT query will use double, but edit control has text in it, and the only way to get that text is to store it into string with the call to GetWindowText.

There is no other way I know if you have a decimal number as edit control's text ( that is why I can not use GetDlgItemInt for example ). Therefore I must perform conversion from string to double.

I hope this clarifies things better.

Best regards.

Sergey Alexandrovich Kryukov 17-Mar-14 17:54pm

Of course, there is no way to get to number, if it is entered by the user, but it does not create a situation where you have limitation by number of decimal positions. You first parse the text into double, and, from this moment, decimals become irrelevant, you have binary double...
—SA

AlwaysLearningNewStuff 17-Mar-14 19:09pm

" You first parse the text into double, and, from this moment, decimals become irrelevant, you have binary double..."

Well, it seems that this is the point where I face the problem.

I believe that database has a fixed range of maximal decimal numbers it can store.

Let us say that range is from -100,5 to 100,5. If user enters 200,5 than an error will occur after I try to execute insert into database.

To protect myself from that, I have used EM_LIMITTEXT to limit the length of text entered text into edit control.

I use strings of fixed length to store the result of GetWindowText.

For example, if the maximal text length is 50, i use wchar_t ExampleString[50] for storing the result of GetWindowText.

If wtof adds a leading zero, and I programmatically load the number into edit control, the text length might be 51 due to the added leading zero. This will cause buffer overflow when I call GetWindowText.

Perhaps my entire approach of getting/storing the text from edit control is bad, but being a beginner I know no better. If you believe this may be a problem, then maybe it is best for me to ask a new question about it.

Best regards.

Sergey Alexandrovich Kryukov 17-Mar-14 19:21pm

"I believe that database has a fixed range of maximal decimal numbers it can store." No! However, I don't know what do you use and why.
Numeric types are binary, they have nothing to do with decimal numbers and they position. This is how you can use floating-point numbers:
http://technet.microsoft.com/en-us/library/ms173773.aspx.
There can be fixed-point numbers, but, by their nature, they are integers. And still, there are no fixed number of decimal places, there is a fixed precision (maximum absolute error is fixed, that's it).
—SA

AlwaysLearningNewStuff 17-Mar-14 19:38pm

It seems that we have managed to establish "a common ground".

I am using MS Access database and I have some fields of type Double to hold decimal numbers.

I am inexperienced, as you have already noticed, so I thought that there was a maximum value that can be stored in a database, and wanted to ensure user does not exceeds it when inputting text.

I guess my previous conclusion was correct, it seems that my approach is wrong. If the length of the number is irrelevant then all I need to do is validate the user input the way you mentioned.

If it will not take much of your time and space, can you please edit your answer to explain me the following:

1. Why double does not have maximum value that can be stored in the database ?

2. Can you recommend me how to change my approach with storing the string from edit control, so I can adapt my code to this new knowledge I have gained ?

If all these are too much then I will post a new question about this, and I ask you to instruct me on which things I should pay attention the most during my research ( I will not ask a question without trying hard to fix the problem myself ).

Thank you for your clarifications.

Best regards.

Sergey Alexandrovich Kryukov 17-Mar-14 20:35pm

Of course double does have maximum and minimum values, and also "delta". The essence is this: there are no two different flavors of single or double (unlike, paradoxically, integer types): these terms imply the same types standardized as IEEE 754: http://en.wikipedia.org/wiki/IEEE_floating_point.

Not only this is one of the most popular structures in all of the computing, this is also one of the least understood. Unfortunately, many "programmers" still don't understand such important numbers as NaNs(Not a Number), plus/minus infinities and the simple fact that floating-point division is a perfectly legal operation...

So, there are no " maximum value that can be stored in the database", there is just one standard maximum value for each of those types.

I already explained how you can change the approach: parse string into double and store that double. There is one problem: I don't know correct parser from string to double for C++ (?!!!). "Standard" parser is a product of extreme idiocy: http://www.cplusplus.com/reference/cstdlib/atof, it returns 0.0 if string is invalid (!!!!!). At the same time, strod for C++/11 is documented to return error in errno: http://www.cplusplus.com/reference/cerrno/errno.

—SA

AlwaysLearningNewStuff 17-Mar-14 22:02pm

I knew about strtod before, and the best thing is that it is even culture dependent!

My pardon, but the IEEE documentation is a bit confusing to me. After going through all those links, I could not find the max/min value, but then again I am tired at the moment so I have missed something.

I have concluded that I do not need to fear from user making an input that will exceed the "maximum double value". I guess it is safe just to take the string, parse it, and if valid just to store it as you have suggested.

The cause of my problem therefore must be my usage of strings of constant length.

Maybe it would be better to just call GetWindowTextLength() and use wstring to create a string of required length? Then I could use GetWindowText without fear of buffer overflow.

Best regards.

It will require for me to change my code but I think it will be worth it.

Sergey Alexandrovich Kryukov 17-Mar-14 22:22pm

The choice wstring vs string depends on the rest or application; numeric formats do not require Unicode, excluding some exotic cultures where those culture-specific numbers (like Indo-Arabic, Easter Arabic, which are rarely used even in their cultures). I don't know why would you worry about buffer overflow at all...
—SA

AlwaysLearningNewStuff 17-Mar-14 22:45pm

"I don't know why would you worry about buffer overflow at all..."

I use fixed length for string when I retrieve the text of edit control with GetWindowText like this GetWindowText( hwndEdit, MyString , 10 );.

Since I have decided to take your advice and just obtain the text and parse it, there is no more need for me to limit the length of the text entered in the edit control to 10.

Therefore if the text in the edit control is longer than 10 there will be buffer overflow.

That is why I ask you if the best approach would be to use GetWindowTextLength() to obtain the length of the edit controls text and then to use wstring ( my project is UNICODE ) to create the string that will store edit control's text.

Then all I have to do is pass that string to strtod.

Best regards.

AlwaysLearningNewStuff 17-Mar-14 23:10pm

I will reimplement my subclassing code the way you suggest, so there is no need not to accept the answer.

Also, 5ed.

All I need to do now is to obtain the string from edit control in a proper way.

Thank you for your patience and help.

Best regards.

Sergey Alexandrovich Kryukov 18-Mar-14 0:32am

Thank you for the conclusions. You are very welcome.

I'll add few words. Real number is a very complex mathematical object which are poorly understood by almost all people. Do you understand that each transcendental number hides any thinkable fragment of any finite piece of information? This is strictly proven fact. And floating-point number is the model of real numbers in the discreet realm.

That is why the notion of maximum and minimum are modeled in a complex way. First, you can consider those number as a (finite) set with order operation defined on it fully defined by a single operator: <= (or >=). From this point of view, the maximum is... PositiveInfinity, and minimum is NegativeInfinity; these two special values can legally participate in comparisons. Note that the set of mathematical real numbers does not have minimum or maximum, of course.

Also, there are "normal" (non-infinite) "maximum" and "minimum" numbers, in C++, they are defined here: http://msdn.microsoft.com/en-us/library/c707ct0t.aspx.

This way, the set with two different "minimum" and two different "maximum" point are used to better approximate mathematical set of numbers which do not have minimum or maximum! How about that?

Now, you should understand that there are two "dimensions" of infinities: infinitely big and infinitely small (in absolute values) numbers, that is "infinitely close" number, defined by the type's "delta". The number is defined as combination of binary mantissa and
binary order of magnitude (logarithm). Maximum floating-point value is maximum mantissa and maximum logarithm. Minimim value is minimum mantissa and maximum (sic!) logarithm. Delta is minimum (with excluded zero) mantissa and minimum logarithm. Aha?

I don't think you need it for solution of your problem, but this is good to understand.

—SA

AlwaysLearningNewStuff 18-Mar-14 0:41am

Although this is too much to grasp, I thank you for the lessons. It is always good to learn new interesting stuff :)

I have made a small demo app to test strtod and validation works.

Instead of using strings of fixed length I am using GetWindowTextLength and wstring so now there is no danger of buffer overflow when retreiving text from edit control.

All I need to do now is to incorporate this in my main application. I think that I can make it.

Again thank you so much, I highly appreciate this. You could have just say "Handle WM_KEYDOWN" and leave me to be poor programmer but you invested genuine effort to teach me the right way.

Hopefully I could one day return the favor.

Best regards Mr.Kryukov.

Sergey Alexandrovich Kryukov 18-Mar-14 0:44am

Ah, I got what did you mean by overflow. This is not exact term, but of course getting the actual length solved the problem...

By the way, I wanted to say that I appreciate your efforts to ask questions in more accurate and comprehensive way, and, of course (especially)... to learn. In contrast to some other members (unfortunately, many of them). Always a pleasure to help someone who can make sure that it would not be a waste of time. :-)

You are very welcome.
Good luck, call again.

—SA

AlwaysLearningNewStuff 20-Mar-14 14:47pm

My pardon for disturbing, but I am really stuck with coloring common controls with Visual Styles enabled[^]. Can you please give it a look and try to help?

Again, I apologize for disturbance, but I have no one else to turn to, and I just can't figure it out.

Thank you.

Best regards.

Sergey Alexandrovich Kryukov 20-Mar-14 15:58pm

No need to apologize; and thank you for the link. After all, if I don't like to answer or cannot, I can simply ignore it. :-)
—SA

AlwaysLearningNewStuff 20-Mar-14 16:19pm

" After all, if I don't like to answer or cannot, I can simply ignore it. :-) "

Indeed, but I still hope we shall "stay in touch". I am also actively searching, and think I have found a concept that might work but will need answers to few follow up questions. If you still decide you want to help leave a comment.

Best regards until then.

Sergey Alexandrovich Kryukov 20-Mar-14 16:30pm

I tool a look — sorry, I'll skip it for now; I did not use the WM_* you mentioned before, and trying out would take too much time. (Usually, I answered based on the code I really tried.)
I'll try to stay in touch and keep myself accessible.
—SA

AlwaysLearningNewStuff 20-Mar-14 16:43pm

Thank you for your time.

" I did not use the WM_* you mentioned before,..."

Do you mean WM_CTLCOLORSTATIC[^] ?

You must have used it before, otherwise how else would you change the color of the text/background of a static control[^] ?

Sergey Alexandrovich Kryukov 20-Mar-14 16:56pm

Yes. No, I think I didn't. Oh gosh, "static contol". The crazy idea of using MFC never visited my head. I never though it was any reasonable way of making UI. Yes, I tried, found it disgusting... I used many other things, including raw Windows API...

Look, controls don't have to use it. You can have light-weight control, all based on WM_PAINT. Isn't it logical?

—SA

AlwaysLearningNewStuff 20-Mar-14 16:59pm

You mean using GDI to draw text instead of static control? That is OK, but what about radio button or checkbox?

You must handle that message in order to get proper text/background color right?

Or is there a better way I am unaware of?

Sergey Alexandrovich Kryukov 20-Mar-14 21:06pm

Yes. And what about radio button or check box? If you need completely new look, you can use GDI with WM_PAINT, painting nothing by default. Some controls allow one or two owner-draw modes. And, rarely, you can pain a little detail on top of default view...
—SA

AlwaysLearningNewStuff 20-Mar-14 22:06pm

I have never done this, but I am interested in trying your approach. Can you clarify it a bit? What do you mean by WM_PAINT?

In order to preserve radio button/checkbox functionality I guess I would need to subclass those controls and handle their WM_PAINT, is that what you say? This would be acceptable solution for me.

If you mean drawing the controls in dialogs/main window's WM_PAINT handler, well I could try that to, but doesn't that mean I would have to recode everything from scratch ?

Thank you.

Best regards.