QTX Framework for Delphi and FPC is available on BitBucket
Jon L. Aasenden - April 13, 2020
Jon L. Aasenden - April 13, 2020
[SHOWTOGROUPS=4,20]
The QTX Framework is a large and mature framework that is presently being organized and open-sourced. It’s essentially 20 years of hard work by me, that I am now consolidating and giving back to the community I love so much.
To make a long story short, I have had way too many hobby projects, to the point where my free time was reduced to zero. So for the past six months I have gradually reduced my life considerably (or the complexity of life), with the goal of just having one hobby (Quartex Media Desktop) and one steady job with no extra side projects.
But, instead of letting good code just sit there collecting dust, why not give it to the community that have shared so much with me? Feels like the right thing to do.
QTX Framework
The first unit to be released into open-source (MPL v2) is qtx.vectors.pas, which gives you four vital class types that have been missing from Delphi (and Freepascal). Classes that are extremely powerful and that should have been made a part of the RTL over a decade ago.
As you probably know, C++ has enjoyed vectors (std::vector) for a long, long time. But I have taken the concept one step further, and coupled it with a unified storage system, so that the vector containers can operate in memory – and with files. So you can now work with large vectors both in memory and on disk. The system is largely expandable, so you can roll your own storage types (cloud, network etc) with relative ease. Same goes for vector allocator classes (for DB ORM style mapping).
The storage system (untyped buffers) deals with pretty much everything you need to manipulate raw, binary data (and then some). And this is where TView comes in, where you can access a chunk of untyped data (regardless of its location in memory or disk) as an array of typed data.
TVector and showing live memory used by the vector container
I published an article on this over at the Embarcadero Community website, which can be useful if you dont fully understand the gravity of this:
Для просмотра ссылки Войдиили Зарегистрируйся
Why should I bother?
Let’s say you have 10.000 Pascal records (or 1000000 for that matter) you want to work with, but you dont want to keep them in memory. Vector containers are now ideal for this line of work thanks to the unified storage system:
If you remove records from the vector, the managed file is physically scaled and truncated
But hey, all of that is easy right? With the exception of live file management you can do this with a TList<T>. Fair enough. But what if you want to look at the data differently, not as TPerson, but rather as bytes or doubles (or another record type altogether)? And keep in mind, the data is managed in a file with no memory overhead – yet the interface remains the same:
Enumerating the bytes in the vector’s buffer, regardless of storage medium
This is just the tip of the iceberg. Vector containers gives you all the perks of old-school “file of record” but without the limitations and problems. Thats just one of many aspects.
Delphi and FPC
Since companies like TMS use Freepascal for their Web Framework, I made sure the code is compatible with both. And there will be plenty of Demos for both Delphi and FPC. Especially as more and more units are added to the framework.
The support for TBit in views is really neat, since you can then access any buffer as an array of bits (which bypasses the problem in C++ containers where TVector<bool> [which is their take on bits in this context] caused some headaches). Heck, the TStorage based classes alone is 3 times faster than TStream, and supports insert and remove functionality (literally scaling the file or memory with truncating if needed), so for binary formats, this should be heaven sent.
Add RTTI and some field-mapping, and you have an ORM engine that will outperform every option out there. Records are faster and more efficient than object instances. And a flat file database is now absurdly simple to make.
Freepascal demo showing how views work, here showing TView
Well, I hope you guys find this useful!
The next unit to be MPL’ed is the parser framework and a ton of parsers for various formats.
Cheers!
[/SHOWTOGROUPS]
The QTX Framework is a large and mature framework that is presently being organized and open-sourced. It’s essentially 20 years of hard work by me, that I am now consolidating and giving back to the community I love so much.
To make a long story short, I have had way too many hobby projects, to the point where my free time was reduced to zero. So for the past six months I have gradually reduced my life considerably (or the complexity of life), with the goal of just having one hobby (Quartex Media Desktop) and one steady job with no extra side projects.
But, instead of letting good code just sit there collecting dust, why not give it to the community that have shared so much with me? Feels like the right thing to do.
QTX Framework
The first unit to be released into open-source (MPL v2) is qtx.vectors.pas, which gives you four vital class types that have been missing from Delphi (and Freepascal). Classes that are extremely powerful and that should have been made a part of the RTL over a decade ago.
- TVector
- TView
- TBitBuffer
- TStorage
- TStorageMemory
- TStorageFile
- TStreamAdapter
- TPropertyBag
- TTyped (*)
- THash (*)
As you probably know, C++ has enjoyed vectors (std::vector) for a long, long time. But I have taken the concept one step further, and coupled it with a unified storage system, so that the vector containers can operate in memory – and with files. So you can now work with large vectors both in memory and on disk. The system is largely expandable, so you can roll your own storage types (cloud, network etc) with relative ease. Same goes for vector allocator classes (for DB ORM style mapping).
The storage system (untyped buffers) deals with pretty much everything you need to manipulate raw, binary data (and then some). And this is where TView comes in, where you can access a chunk of untyped data (regardless of its location in memory or disk) as an array of typed data.
TVector and showing live memory used by the vector container
I published an article on this over at the Embarcadero Community website, which can be useful if you dont fully understand the gravity of this:
Для просмотра ссылки Войди
Why should I bother?
Let’s say you have 10.000 Pascal records (or 1000000 for that matter) you want to work with, but you dont want to keep them in memory. Vector containers are now ideal for this line of work thanks to the unified storage system:
If you remove records from the vector, the managed file is physically scaled and truncated
But hey, all of that is easy right? With the exception of live file management you can do this with a TList<T>. Fair enough. But what if you want to look at the data differently, not as TPerson, but rather as bytes or doubles (or another record type altogether)? And keep in mind, the data is managed in a file with no memory overhead – yet the interface remains the same:
Enumerating the bytes in the vector’s buffer, regardless of storage medium
This is just the tip of the iceberg. Vector containers gives you all the perks of old-school “file of record” but without the limitations and problems. Thats just one of many aspects.
Delphi and FPC
Since companies like TMS use Freepascal for their Web Framework, I made sure the code is compatible with both. And there will be plenty of Demos for both Delphi and FPC. Especially as more and more units are added to the framework.
The support for TBit in views is really neat, since you can then access any buffer as an array of bits (which bypasses the problem in C++ containers where TVector<bool> [which is their take on bits in this context] caused some headaches). Heck, the TStorage based classes alone is 3 times faster than TStream, and supports insert and remove functionality (literally scaling the file or memory with truncating if needed), so for binary formats, this should be heaven sent.
Add RTTI and some field-mapping, and you have an ORM engine that will outperform every option out there. Records are faster and more efficient than object instances. And a flat file database is now absurdly simple to make.
Freepascal demo showing how views work, here showing TView
Well, I hope you guys find this useful!
The next unit to be MPL’ed is the parser framework and a ton of parsers for various formats.
Cheers!
[/SHOWTOGROUPS]
