Friday, December 3, 2021

The trap of wanting to make it perfect, or #D.MVVM what takes so long?

Hello, my friends!

It's quite some time since my last blog post - I know - it was/is a busy year. 



I regularly get emails asking when my MVVM framework will be ready for shipping. This is a very good question... (It was ready nearly two years ago - my designated release date was 2019-12-06)

What? 

Yes, I'm sorry this is the sad truth because just one week before my estimated release date of the beta version I decides to cancel it.

Why?

As it is/was a source code distribution, after a close look at the code, it was messy and really unreadable. 

So I started a little bit of refactoring.

To demonstrate the power of the framework I also started writing some sample applications for VCL and FMX and also some mockups for unit tests.

Getting more and more into real-world design details, everything was working so far, I found some places that could be better, or in other words: The comfort of the framework was not easy enough to create some special bindings. 

Of course, from the early beginning, the main binding unit was able to bind everything (by "Hand", I mean in Code).

At this point, I could just follow the bad rule: "It compiles, It works, just ship it". 

But I expected more, so I deleted all binding rules and the main binding unit. There has to be a better way. (At this point I also decided, not to stick as close as before to the MS-C# MVVM examples)

Just imagine: You have a ComboBox for "Mr., Mrs, Sir, Dear" this might be a close list (Style: csDropDownList) or just (csDropDown) where you are able to put in a free text.

If you didn't watch any of my MVVM-Youtube Videos. My Framework could bind to "normal" properties at the ViewModel, but I also created my own TProperty<T> fields which are much better than just properties. E.g. all the PropertyChanged or INotifyPropertyChanged events are builtin.

So if you want to set the Item list from the ViewModel (like reading the Items from a database), you probably want to bind to the Itemindex and to the Text-Property and perhaps (on FMX) to the Textpromp property. The UI often has to enable the controls or even set the visibility and last but not least perhaps you want to set the hint text. (In most cases you want to handle the OnChange event or the OnClick)

In the MVVM-Pattern you normally do not set a Visual Control Visibility in the ViewModel, or even the color. This is part of the View on some conditions. Like: ViewModel.CanSave or ViewModel.CriticalState (to show a warning label in red)!

After talking to many Delphi Developer I decides to enable the visual settings also to the TProperty<T>. So if you have

Name : TEdit;

on the View (Form) and a property field like:

fName : TProperty<String>;

you can write

fName.Enabled := false;
fName.Text    := 'Frank';
fName.Hint    := 'could not be changed';

in the ViewModel.

If you think that is not like MVVM should be... Sorry, I don't care!

So it is even possible to Name it EdName:TEdit on the View and set your own naming converter to bind all TEdits with Ed[Name]:TEdit as a prefix to f[Name]:TProperty<String>.

So back to our TComboBox problem:

With

Salutation : TCombobox;

the ViewModel could look like this (Property Version):

{ I skipped the getter/setter here to make the code more simple }

TPersonViewModel = Class(TViewModel)
  private
   // ...
  public
    Property Salutation : String;
    Property SalutationItems : TStrings;
    Property SalutationIndex : Integer;
    Property CanShowSaluation : boolean;
    Property CanEditSalutation : boolean;
end;

Or mixed with command Methods like:

TPersonViewModel = Class(TViewModel)
  public
    procedure SalutationClicked;

    function  CanEditSalutation : boolean;
    function  CanShowSalutation : boolean;
 
    Property Salutation : String;
    Property SalutationItems : TStrings;
    Property SalutationIndex : Integer;
end;

Or with my TProperty<T> fields:

TPersonViewModel = Class(TViewModel)
  private
    fSalutation      : TVisualProperty<String>;
    [ OnValueChanged('SelectionChanged') ] 
    fSalutationIndex : TProperty<Integer>;
    fSalutationItems : TListProperty<String>; 
  protected
    procedure SelectionChanged;
  public
    procedure SalutationClicked;
end;

My goal was/is: To support all these different methods and also bind everything without even one single line of code. Don't get me wrong - also nothing set in the Object Inspector. No double click on an event! Just a barebone Form clicked together nothing more.

That's why I had to reinvent the binding rules syntax!

So one ComboBox must auto bind by name (perhaps with a rename rule) to three different fields! Only one should have the visual parts (that should raise an error if more than one is bound). On value change should automatically call the SelectionChanged procedure, the Index property field should be an Integer nothing else, and the OnChange, OnChangeTracking (FMX), and the OnClick event should be connected to the corresponding fields, properties, or procedures. You should be able to mix my Property-Field with normal Properties. 

On top:

There is one visual component with one OnChange event... Where to bind?

of course to fSalutation:TVisualProperty<String>

and not to the two others. So the syntax for the binding should be able to define where to bind the events...

For the Combobox, the style like 
csDropDownList limits the possibility to bind the control to the string property and in this case the OnChange event should bind to the 

fSalutationIndex:TProperty<Integer>
 

field. (Or just to a SalutationChanged method...)

Not an easy task you can imagine!

I have done some work over the last two years, but the main part of the binding stuff I didn't touch. Or even worse, I no longer understood my own source code. 

I tried my best, but without renaming methods and variables it was impossible to read. I often looked for a distraction so that I would not have to deal with the actual problem. It certainly took me 10 attempts to finally start developing the binding routines again. 

So...

There is good news and bad news!

I think I see light at the end of the tunnel, but I need a few more months to get an alpha version on the road. (Ahh and btw. I developed an IDE-Wizard to create a new #DMVVM application for VCL and FMX also not 100% finished and not tested with D11, yet).

I'm still in my own trap to develop the best MVVM-Framework money can buy...

How to bind Stringgrids or event TreeViews are totally beyond the scope of this blog post...

So please stay tuned...


 

Wednesday, September 22, 2021

US Version of my German CodeRage 2021 session is online!

My contribution to the German CodeRage 2021 on September 9, 2021.

The 40 tricks/techniques that every developer should know.

This is not about knowing the different software patterns, but the little tricks that often make development easier. In this session very simple things are shown, but also quite sophisticated techniques are demonstrated. So there should be something for the Delphi beginner as well as for the advanced Delphi developer.

It's now online with the US-Translation on my YouTube Channel!


Sunday, September 5, 2021

CodeRage Germany Sep 9th, 2021

Of course, I'm doing a session on the german CodeRage 2021. My session is not about FMX and not about MVVM... Hard to believe, isn't it?



This time my session topic is:

The 40 tricks/techniques that every developer should know.

This is not about knowing the different software patterns, but the little tricks that often make development easier. In the session very simple things are shown, but also quite sophisticated techniques are demonstrated. So there should be something for the Delphi beginner as well as for the advanced Delphi developer.

As this is the German CodeRage my session is in the german language. But if I find the time I will provide an English Version on my Youtube Page - Delphiprofi.




Thursday, July 22, 2021

What happens when the flood comes?

Maybe you have seen it in the news or are affected yourself...

There was a flood catastrophe in Germany, due to heavily rained and overfilled dams.

But that's not what this blog post is about. 

The question is: How well is your network/server infrastructure set up to function again as quickly as possible in such a crisis?

I'm not talking about backups here, everyone should have those anyway! I'm talking about the structure and the software you use.

A small example: 

Is the MySQL server IP hardcoded or is it stored in an INI-file? Is a URL used or a domain controller? 

Is internet access needed for smooth functionality, because the server may have to synchronize data with an external server?

What if you change IP addresses due to an emergency solution?

I just imagine how your thoughts circle over your own network structure and creates an OMG...

What I've learned in the last few days:

  • It's just terrible how dependent we are on the Internet these days.
  • It must be possible to replace ground Internet access with a mobile solution at any time.
  • Document folders and paper storage are good. (Still necessary in Germany, unfortunately). But you can't digitize too much! Especially when you're looking for an insurance contract and can only wave behind the corresponding folders as it just disappears into the floods.
  • A pile of paper according to the motto: I'll scan that when I have time for it... Will never exist anymore.
  • Software that needs a fixed IP must at least be able to load this information from an INI-file.
  • Software that can only handle IP4 must also be able to handle IP6.
  • Software that gives permissions based on the Mac address needs to be redesigned.
  • Software that requires a domain controller or a fixed network name must be redesigned.
  • If a local server is needed on the network, it must be possible to replace it with a cloud solution.
  • Password storage software that encrypts all collected access passwords must have a simple memorable password to open the information!
  • IP telephony is crap.
  • If there is no electricity and no internet, all-important numbers must be stored in the cell phone. (Preferably also cell phone numbers because your opposite surely also has IP phones that do not work without internet and without electricity.

Well - this is all doable. And I'm sure I will find the time to do this ASAP.

We were lucky this time. No employee is injured and everything else is replaceable. Unfortunately, there are far too many who have been hit much harder. My thoughts are with all those who lost much more or everything.

Saturday, June 19, 2021

Is HTTPS enough to secure your connection to a Webservice?

For many years, hardly anyone has bothered to secure their own server with a certificate or HTTPS. Why?

Because certificates are expensive and there was no reason to secure a private handwritten HTML page with the family pictures.

Bussines Websites with money transactions were the first to go HTTPS.

In the old days, many or nearly every MySQL Server had an open port to the internet. Today you can't do that anymore.

Is it so? Everybody will probably answer this question with:

"You have to connect to the database over a REST-Service, for safety!".

But what is the difference between login to a REST-Service and login to a Database-Server? Log in to a Database-Server is done with User and Password. And to a WebService? Normally with basic authentication, what under the hood is User and Password.

So why is it better to use a Webservice instead of a direct connection? The WebService could do more checks. Well, besides the security, if you send a request directly to your database server everything has to be transferred to the client. Inside a Webservice, you can prefilter manage and do different things before you send the data to the client. This could reduce the amount of data that must be transferred, but on the other hand, if you use JSON the data gets expanded by all the fancy { and } and some data gets really big, like pictures or other data blobs, because every byte has to be encoded with Base64 or Hexstrings.

As the internet gets faster and faster it looks like nobody cares anymore about the amound of data. The days for handwritten, optimized HTML are long gone. Everybody is using some kind of framework that transfers big javascript files with every response.

So, let's say that more bytes are transferred, so the database connection is certainly slower, but nobody cares about it, because of the increased security.

If you are using HTTP everybody in the middle could read all your data, that's why you have to use HTTPS. Thanks to Let's Encrypt it is not a factor of money anymore. But how secure is your connection?

If you open a website with your browser, you get this little lock in front of your URL. Nice we are safe! Are you sure? Let's leave that aside for now. What about when we query a REST(Full) web service from within an application?

We use a component or native HTTP-Get to the URL of our web service. The URL has HTTPS at the beginning.

Are you sure you are connected to your server?

Let's check if the certificate is correct! That's a really good idea. So we write a little proof of concept application. Intercept the certificate-check-call and compare if the information matches our certificate. 

And?

Fail! This is not our certificate! Is there a Man-In-The-Middle? Yes, it's your virus-scanner application! A very bad habit of a virus scanner is to check all HTTP and HTTPS traffic to find bad HTML that could harm your PC or your data.

Of course, it could also be another tool, such as Fiddler. But that should be clear to me because in this case, I started the tool by myself. However, on the way from my internet connection to my server, a corresponding tool could also be used.

I have no clue if the virus scanner has a backdoor and I'm not an aluminum hat carrier. But let's assume for the moment that this is the case, then the login data to your web service has been sent directly to the NSA in plain text. LOL...NSA in this case stands for all strangers who should not receive your data or logins.

If you have to protect your data stream, in this case, you have two options. 1.) Ignore it and may the NSA be happy with my login information, or 2.) Stop the data transfer.

So we are doomed?

There is one possibility you can do. (Not so easy on a website, but from your application).

Crypt your traffic!

Use the public RSA-Key of your Service to crypt your request. Send the public key of your application to the server so that the server could encode the data for you. Perhaps change the key for every new request.

Doesn't that sound familiar? 

Yes, it's like the SSH/HTTPS handshake. Thanks, virus scanner developer. 

Oh, by the way, if you are doing the crypting by yourself anyway, there is no need for an SSL certificate in the first place.

One nice side effect: Any tool that is trying to read your data stream can not fool you with its own certificate to read the data.

If you have read this far, I have one last question: Should I do a webinar / Youtube video on this topic?


Please leave a comment.






Wednesday, May 26, 2021

Stackoverflow: This time with Delphi in the list.

There is an annual developer survey going on on Stack Overflow. Delphi is in the language list, now we need Delphi developers answering the survey and ticking Delphi to show it is relevant. Stack Overflow account is not required.

So please fill out the survey and on the environment question, please enter "RAD Studio" in the empty field.

Here is the link.

Monday, May 3, 2021

TCoffeeAndCode - Threading and Concurrency

Hello!

Are you interested in learning more about threading and concurrency?

On Wed, May 5, 2021, 7:00 PM - 9:00 PM CEST 

or for better understanding 17:00-19:00 UTC,

you can listen to how we approach these things and what our opinions are on these topics.

Here is the replay!

Saturday, February 27, 2021

Delphi 10.4.2 - What is your current Hardware for Development?

Well, this is of course not related to this newest version of Delphi, it's the right question for every Developer, but with every new version of Delphi, this topic is back on the table. If you're looking for my opinion on 10.4.2 - please just scroll down!

Declaimer: This is not related to Embarcaderos specifications of minimum requirements, this is "just" my point of view.


Let's start with a little example:

If you were a carpenter, would you buy a jigsaw from the hardware store for 49.95 or would you have a high-end circular saw table that can do everything you need?

If you have to cut down an entire forest, you wouldn't use a handsaw, but a chainsaw, would you?

Is it possible that your 3D printer cost more than the laptop on which you "work"? 

I just read a question about the new IDE: "If I press a key and the LSP has to update, I can hear my Laptop-Fan is ramping up... 

Laptop? Really? Are you kidding me... 

There are some Laptops out there, you can probably use...But for this price, you better buy a third Monitor.

Ok, If my system is mining in the background and I'm watching a Video while loaded up 3 VM with different IDE's sometimes I hear the fans of my watercooler spin a little bit faster. Just joking, minding on the GPU is stupid.

Your Workstation is your main tool for work, yes I call it Workstation, Powerhouse, whatever... If you call it "My PC" and you are not able to run the latest AAA games on it. Buy a new one!

After 35 years of Development here is my list of NO-GOS (in unsorted order):

  1. Use a non-SSD Harddrive or an SSD attached via USB.
  2. Not installed the latest OS
  3. less RAM than 16MB
  4. Screen resolution less than 1920x1080
  5. Only one Monitor.
  6. Not using a VM
  7. Using a non-standard keyboard. (perhaps you can work on a Laptop keyboard I could not)
  8. Not using a Source-Code-Management System (svn, git, mercurial)
  9. Firemonkey App development without my FDK... ;-)
So what are my recommendations you ask?
  1. I would never want to develop without a VM again. So: Use a VM. I like VMware, there are some issues with it but overall it runs perfectly. Using a VM set the bar for a good system a little bit higher! But having auto-snapshots every hour and be able to revert the installation back to a working copy if you have installed something stupid or your installation just broke down is worth it.
  2. Your CPU should have a core-count that is double as high as you would like to have in your windows system. I prefer 4 cores so with hyperthreading you have 8 logical cores in my windows. That's why my CPU has 8 cores with 16 logical cores. 4 for the VM and 4 for the Host. I'm an Intel-Fan-Boy, never tried AMD. My colleague has bought a board with the latest AMD CPU and he has so many problems...
  3. Use 3 Screens. 
    1. Center screen for the IDE. 
    2. Left screen for the undocked the Objectinstpector and structure view. 
    3. Right Screen for the program group and the component palette. 
    4. My Line for breaking up the source line is not at 80 anymore, it's on 160!
    5. While debugging, place your program on the right screen (program group not shown). 
    6. Most of the time when doing boring stuff I put my OI docked at the left side of the IDE on the center screen because I switch my left screen to my Host for watching YouTube videos or other stuff...

  4. If you want the "same speed" inside a VM that you would expect from your main system, Harddisk and RAM performance is the key.
    1. I've installed 64GB of RAM so I can start 3 VM's with 16GB and still have 16GB left for the host system. Or any other combination. 
    2. Perhaps you think: RAM-Speed is not so important. I did not test this and perhaps high-speed RAM has a not-so-big impact on your system. I prefer RAM modules that are recommended by the Mainboard Manufacture and often higher speeds come with a huge impact on the price.  But this looks not so bad compared to the rest of the world.


    3. On the Drive-Side there is no room for low-cost at all! Maximum performance is the key to a smooth-running VM because the VM is often very disk intence.
      1. You get the best performance with an M.2 Drive. If you combine 2 or 4 of these high-speed M.2 to a RAID 0 stripe-set you are pretty close to the maximum speed your bus could handle.
      2. Never ever use your VM on a Harddrive or a single SSD connected to a SATA connector. I had SSD failures in a RAID 10 and this saved my installation.
      3. For Backup I have a RAID 10 of 4 SSD's and a RAID 5 with 3 M.2! Here is the trick: I use the hourly auto-backup of VMWare. That's fine. From time to time I shutdown my VM and copy it (1,6 TB) from a RAID 0 (4 x M.2) to to the RAID 5. This is the best performance I could get. After a few minutes, I can restart my VM and then copy the files for this RAID over USB C 3.1 Gen II to an M2. in a fan-cooled external housing. With this, I get the maximum performance USB could handle. I really would like to have a Thunderbolt connection (should double the speed), but the card I bought was not working and I have no spare slot on my Mainboard. 

  5. I know some people prefer Air-Cooled systems. But I'm using a custom loop for so many years now and the only problem I had once was because my soft tubes are getting old and one fitting was leaking. My System runs quiet, always below 50° C (often below 45° C) and with adding a little bit of water from time to time I had no maintenance to do. (Yes, I just use water, not any colorful additive or even distilled water) That's why I have no little particles clocking up my block. With my last CPU/Mainboard update I installed a new reservoir (the old Plexiglass had got some cracks) and switched to a new pump (just in case). This was the only maintenance in 10 years. 
  6. Install a really good graphics card. Perhaps not so important for VCL developers, but as FMX uses the GPU to render, you'll get the speed you've never seen before with VCL.
  7. For the App development, I'm using a Mac-Mini. Ok, I switched out the hard drive and installed an SSD. (I really should test a macOS-VM sometime). 
  8. And similar to point 9 of my no-gos - use my FDK for your FMX, VCL, and App-Development on the mobile platforms.

Anything else?

hmm... I'm still on D2007 with my main application. (Not finding the time to fix all the Unicode warnings). But the old compiler is also able to compile more than 2. Mio. LOC in under 16 seconds for a full build.
 
Perhaps you call my recommendations overkill, I call it: Best bang for the bug... I spend so many hours at my workstation and I have no time to wait for something to finish. That's why I want the best performance you can buy. (For a reasonable amount of money).

I love the statement about electric cars from the best rally driver of the world Walter Röhl: 
I'm too old and I don't have enough time left to wait at the loading dock. (Translated)

In my case: I'm too old to wait for the compiler. If I currently working in a unit to implement something and I have to take a look into a different unit (normally I have too many open units, often more than 20), it is faster to hit compile and let the compiler find the position I am working on, then to search for the unit or ever do a text-search inside the same unit... ;-)

So - don't complain that something is too slow, beef up your system!

If EMBT tries to speed up the IDE and the compiler at the same time, we will get a faster and faster development environment or at least a constant speed as the possibility grows.

Back to 10.4.2


In my tests 10.4.2 is faster than 10.4.1 and 10.3.3. There is of course an impact now many components you have installed and how long your search path is. Also, there are differences when compiling inside the IDE or using the external compile or MSBuild.

Besides the numbers, the IDE feels snappier and there is a lot less flicker (or nearly/no flicker anymore). There is a little loading screen, popping up when the IDE is loading units. You can disable this, but for the first time you get a notification of what the IDE is doing - I love it.

There is also this new feature of the LSP. You know the old red wriggling lines below errors that were wrong in most cases! With the new settings, you could switch to different lines or dotted lines that also could have different colors. Blue, Yellow, and Red for "not used", "hint" and "error"... My first reaction to this was...

Why on earth did they do this?

But while using this feature for some days - I love it - you can faster correct your mistakes or change a var name because you misspelled it.  This allows you to develop faster, believe me.

Overall it feels better - yes there are still some little bugs in the IDE - we all know, but I can live with the new version!

And as always:
We hope for the next version (10.5) and that the little bug gets fixed!

So long, have fun with the new version.


Friday, February 5, 2021

My road to a useable MVVM Pattern implementation for Delphi!

If you read my blog regularly, the MVVM topic will be nothing new to you.

But please take your time and fill out my new survey, thanks for that, of course, even if you visit my blog for the first time. I would love it if you can also be part of this survey

If you are a first-time-reader... Welcome to my blog. Please read the other posts about my #D.MVVM implementation. As you can see, this subject is not new to me and my knowledge has, of course, improved over the years. (I hope, I'm right with this statement)



Here are the links to my other MVVM-related blog-posts in reverse order:

There was a blogpost from Oct. 2019 with the question: How long does it take to develop a "complete MVVM framework for Delphi. (Link above) the answer is: Not long if you take shortcuts or too long if you are a single developer with no help. But I won't cry, because my goal was not to do it the easy way.

If you do not make it down to the end:
Here is the link to my new survey!
But perhaps you want to know why I've created a new one? ;-) 


Let's collect - one more time - my goals:

  1. The framework must be as fast as possible. (My goal is - it is supposed to be the fastest MVVM framework available for Delphi.)
  2. The development time should not be significantly longer than using the RAD-Approach. At best, the development time should be shorter, comparing to a non MVVM, but not "everything in the Form unit" - development.
  3. Everything that could be done in a background thread should be done in a background thread.
  4. Support for FMX and VCL. Including the "special needs" for the FMX UI.
  5. It must definitely run on ALL platforms.
  6. 100% Unit-Test-able of your Application. (Switch for disabling the background thread to do a better Unit-Testing).
  7. Mockup's for View-Components to test the Binding of your App.
  8. A nice Wizard should be installed into the IDE to create a new VCL or FMX-MVVM Application. This Wizard should be changeable over an Ini-File to make a User-Configuration without recompiling the Wizard-DLL.
  9. One central binding point (Composition.Root) 
  10. If you are an FDK-User (My Firemonkey-Development-Kit) the framework should work streamless as a plugin, with all the benefits of my ORM-Lite, threading, or Client-Server units.
  11. No need for special components. (Components must work without an extension)
  12. Open to User-Implementation for non-standard components.
  13. The View (Form) should not have a single line of code. Or should not need any code-behind...
  14. Visual Livebindings must not be used.
  15. A developer who has NEVER dealt with MVVM has to get along with my framework IMMEDIATELY.
  16. A good developer who has already used MVVM should find his way around immediately. (Except for some things that I have implemented differently than MS).
  17. We are Delphi-Developer - If we are doing MVVM differently than our C# friends - that's fine!
  18. Views must be able to be composed of any number of sub-Views (frames) without writing a single line of code for them. (Sub-Views must be interchangeable for special requirements)
  19. It must also be possible for a sub-View to be a ViewPort3D (FMX-Only). Full support of 3D objects from the ViewModel will be another thing for further investigation. This could be very interesting for Game-Development, but that is completely another topic.
  20. User-Level and flagged Menu-System with auto Action-Binding.
  21. Multiple View for a ViewModel, also with sub-View/ViewModel Binding.
  22. Conventions over definitions.
    1. This is especially true for automatic binding by naming convention. Example: Name: TEdit should auto bind to the fName: TProperty<String> or to a normal property Name: String. The Binding is connected to the Text-Property of the TEdit Component and the OnChange and OnChangeTracking (FMX Only) event should update the ViewModel.
    2. User-defined naming convention... Bind Name:TEdit to fEdNameField : TProperty<String>
    3. Global naming rules and also naming rules by TypeInfo.
    4. Overwrite any predefined rule with an attribute.
  23. Multi-Binding to different properties.
    1. Example: Save: TButton.OnClick -> Procedure DoSave
    2. Save :TButton.Enabled -> fCanSave : TProperty<boolean> or Property CanSave : boolean
    3. Fone : TCombobox.Items -> fFoneItems : TListProperty<String>.
    4. Also user-defined naming: "Can"+Save;"Do"+Save could be changed to: e.G.: -> "CheckFor"+Save;"Execute"+Save.
  24. Optimal performance on ".Strings";".Items";".Lines" - never ever send the complete list on a single change. A virtual created adapter must take care of the updates. (e.G. for a single changed line in a Memo of 10.000 lines it is a nogo to update the Text-Property of the Memo.)
  25. Data-Verification on Property-Level to check the fields inside the ViewModel.
  26. Listview and Listbox special treatment for FMX.
  27. A data model for VCL Grids - The FMX Grid already has a data model.
  28. A not so easy implementation of an MVVM-Treeview. (Perhaps not in the Beta)
  29. Complete logging of all internal functions should be possible. (Only for the Plugin Version)
  30. Error messages that are normally written in English must be able to be translated by an external language file.

The MVVM-Pattern has the Model included in the Pattern-Naming, but actually, the Model is not really a part of the Pattern. But in my case:

  1. Only if the Framework supports the transfer of all the collected data from the ViewModel to the Model, my goal of: "As fast as RAD Development" is possible. 
  2. The Framework must have a Datacontext that can make the transfer from the ViewModel to the Model, and back.
  3. The Datacontext should be able to convert, combine and separate fields as they are transferred back and forth.
  4. The Model should directly work as an ORM or REST-full Web interface. 
  5. As an FDK-Plugin it should also work with my Client-Server-REST-Units.

So far so good...

There is light at the end of the development tunnel! I'm really happy with the actual implementation. I think I have done all the necessary improvements for the first version. and - as always - I would have programmed the framework even if I wasn't planning to offer it commercially. Of course, programming is a bit different when the source code is handed over to other developers. Certainly, I could have made a lot of things more static or simpler.

And now?

There will be a Beta-Version and an Early-Bird-Order! But maybe this time I should do an Alpha -Version too - if I can find developers interested in helping me finish the release version.

But how and why?

After years of development - only in my spare time, of course - perhaps I've done something wrong, perhaps I should rename a method or change params of a function. Whatever... I don't want to make breaking changes from Beta to Release.

I know - some of you - are waiting for the release of my Framework, but besides the old survey, I don't know how many Developers would still like to be part of the Alpha or Beta-Team. (or even would like to buy my #D.MVVM-Framework). (Old survey filled by 177 Users).
I've talked to some of these developers and they can't wait to get started with my framework. (I'm super happy to hear that, of course).

A comment in advance from René Höger:
"Is a lot of information! Those who have dealt with this, I think will be hot on it! I don't know what your feedback is so far, but I think this could be a milestone. When I think of all my old applications! But maybe I am biased because I have already received an introduction into your framework!" 

💖 Thanx René!

Perhaps I need to write some kind of NDA contract (never done before). One thing I want to avoid, after all these hours, is to find my Source-Code as a zip on a hacker site.

Before I could start the Alpha-Phase, I have to:
  1. Open up a ticket system or some kind of bulletin board?
  2. Perhaps a discord server?
  3. Perhaps a slack group?
  4. Think about how to collect User-contributions...
  5. Think about contribution benefits?
  6. Late or discount payment for Alpha & Beta users? 
How do I distinguish between a developer who "just" wants to see cool source code and a developer who wants to actively contribute to the framework? Maybe I should just assume the best and trust that Delphi developers are all super nice people who are interested in using this technology. I think I'll stick with that idea or a very little "fee" just to check if the developer really wants it? Perhaps some kind of reward system?

If you've made it down here - Thanx for reading and please be so kind and fill out my survey!


Saturday, January 23, 2021

The IFDEF Problem!


If you are using {$IFDEF} in your source - normally it's fine...

e.G.

{$IFDEF DEBUG}
ShowMessage('Foo');
{$ENDIF}


Why, because this DEBUG is handled by the IDE.

But what if you want to use your own?

{$IFDEF IMPORTANT}
Result := 42;
{$ENDIF}

If your application depends on this, you always have to do an else part to make sure the Name "IMPORTANT" is set.

{$IFDEF IMPORTANT}
Result := 42;
{$ELSE}
! Too stupid set this!
{$ENDIF}

I don't want to have this in my source in every place I just want to exclude or include something.

Like:

{$IFDEF UseLogger}
Log('Ready to run');
{$ENDIF}

In this case, I want to be able to include logging, but on the other hand, I want to be sure that I disabled logging for a good reason, not just forgot to set the {$DEFINE UseLogger}.

That's why I like to use const's:

Unit Defines;

Interface
  Const
    UseLogger = true;
Implementation

end.

{$IF Uselogger} is also not working if you just forgot to "uses" your Defines.pas Unit.

That's why I'm using the not implementation.

Unit Defines;

Interface
  Const
    NotUseLogging = false; // means use logger
Implementation

end.

Because using:

{$IF not(NotUseLogging)} has the benefit of checking if there is a symbol named NotUseLogging!

So if you forgot to use the Defines Unit you are getting an error at compile time. I'm using the approach for my FDK and also for my #D.MVVM Framework.

This works fine... But there is one drawback: You are unable to change these settings per project. In the past, I've changed the source of my defines unit, but changing this back and forth is a pain. Of course, this unit is named Delphiprofi.FDK.IFDEF.pas!

The other drawback is, you are unable to use the {$UNDEF Foo} in a Unit. So what is the best solution?

Let's collect the necessary things we want to accomplish:

We want to be sure to set important things globally for the whole framework. That's why I use the {$IF Not(Name)} statement. In the past, we all used this stupid {$I defines.inc} approach, but I hate that so much and it also has the problem that your {$IFDEF Foo} is unable to check if you just forgot to include the file. How often have you searched for a problem, because you misspelled a conditional define? I always type  {$IFDEF WINDOWS} instead of   {$IFDEF MSWINDOWS}.
Of course, we can still use normal {$DEFINE Foo} and {$IFDEF Foo} for local unit settings during development. While using the "if not trick" in every unit we are sure that all {$IF not} statements have the right setting... 
And how to use these settings on a project level? Of Course with a normal {$IFDEF} in the Delphiprofi.FDK.IFDEF.pas Unit. In my case, I take this idea one step further by using a class like:

ExcludeFDK = class sealed 
  public
    const
     Logging     = false;
     //... more in the original source
end;

With this, I just can type ExcludeFDK "." and the IDE gives me all possible settings (with the right spelling).

The If looks then like this:

{$IF Not(ExcludeFDK.Logging)}
FDKConfig.GetLogger( Self ).Warning( 'This (%s) should not be happend',[aMSG]);
{$ENDIF}

btw. The FDK is also able to disable logging, based on the Type of the Class the logger is fired from. If I want to disable all the logging from my DI - Container, I just set the exclude of this class in code. Very handy!

To be able to include or exclude - in this example - the logging from a project I can use normal defines on the project level. In this case, it is safe, because I know that all units are set correctly. This it how it looks:

ExcludeFDK = class sealed 
  public
    const
     {$IFNDEF FDKNoLogging}
     Logging     = false;
     {$ELSE}
     Logging     = true;
     {$ENDIF}
     //... more in the original source
end;

If I have no conditional defines in my project settings, the behavior is the default. But by using FDKNoLogging all the logging functionality is not linked to the project.

But this is only necessary if I want to exclude the logging completely from a project to reduce the size. If I keep the logging linked, but do not set a logging target, a NIL logger (empty procedure call) is called with the least possible CPU cycles. So you are able to activate logging - for example - over a command line param.

If you want more information about the logging features like Log to Console, Debug Window, Testinsight, SmartInspect, or to a remote server or any of the other targets, please leave a comment.

That's it for the moment... More is coming...

Friday, January 15, 2021

JWT - The JSON Web Token!

Sometimes I use my free time to discover something new (or new for me).

This time it should be the JSON Web Token.

After the first consideration, I didn't understand the use of it at all... 

I (the server) create a token (Header. Payload), make Base64 out of it, and add a signature by hash with a key. 

OK, so far so good. If I send this token to the client, the client can take a look at it with a Base64 decode. 

(Why should I allow the client to do that?).

The client should send me this token with the next request and I (the server) can apply the hash again and check (the client doesn't know my secret key, which I used for the hash), if the hash matches, that is the way I can check, that it is still "my/valid" token. With the payload, I can then identify the user.

Fine - where is the use case?

The server generates the token after the user-login and doesn't need to store this session token. Cool.

How else was it done in the past? 

Generated a GUID, add a few random bytes, then maybe a cipher, and put this into a database. Next time you could just access the database with the token and after the access, you had everything you needed from the user. And the Client does not know anything about the token or what is stored inside - I prefer this, but:

With the JWT, the session database is not needed and the payload of the token can be used to load the user information. Perhaps for your web application, this simple "User is Valid" information is enough! 

The structure of the token is like this:

Base64(Header) "." Base64(PayLoad) "." Base64(Hash("Header.Payload",Key))

Header and Payload are JSON strings. Of course, someone has made countless thoughts about what has to be contained in the JSON. 

So the header could look like this : {"type": "JWT", "alg": "HS256"}

Token type and hash algorithm.

There is also a rule for the payload (RFC 7519). This may all have a justification and looks like this:

{"sub":"Frank Lauter","iat":1610651797,"exp":1610738197,"iss":"FDK-Token"}

Subject, Issuedat (Timestamp), exp (Expiration Time), iss (Issuer) e.g. the library used.

The algorithm works of course also, if you write completely different JSON in there.

One trick you have to keep in mind is that not all characters of the Base64 encoding can be used as URL-Querry-Param. Normally you would think that just a URLEncode is called because it makes the usual %HexValue stories, but this is not so and also the "=" at the end must be deleted. 

Facebook also uses this coding... As Marco wrote 2012!

If you google for "JWT and Delphi" the first link will be a github repo. 

These 24 units - 173KB source - work very well! ;-)

Perhaps you will like the source and if you have to address all the RFC-Rules you might want to use this... 

Or you just write this few lines in your Webbrocker application:

Yes, you have to insert the DateTime JSON Value, and I set it fix to SHA256 bit. 

You need


If you get back the token with the next call you just use:

for this you need the URLDecode:

You can of course improve this little demo and support different Hashes, or use your own content and give a shit about the RFC's - My Server, MyApp, My Rules!

Have fun...