-
Content Count
302 -
Joined
-
Last visited
-
Days Won
5
Die Holländer last won the day on March 12
Die Holländer had the most liked content!
Community Reputation
93 ExcellentTechnical Information
-
Delphi-Version
Delphi 12 Athens
Recent Profile Visitors
-
No, 1-2 years.
-
New Delphi features in Delphi 13
Die Holländer replied to David Heffernan's topic in RTL and Delphi Object Pascal
If FMX is mentioned in the prompt: Check if FMX or VCL has a TagObject property. Provide Feedback FMX has a TagObject property. Specifically, the FMX.Types.TFmxObject class includes a TagObject property, which allows you to store any TObject value. VCL, on the other hand, does not have a TagObject property. Instead, VCL components typically have a Tag property, which is of type NativeInt and is used to store integer values. For more information on the TagObject property in FMX, you can refer to the official documentation: FMX.Types.TFmxObject.TagObject. -
suggestion for 2 new su forum: AI usage and AI coding
Die Holländer replied to Javier Tarí's topic in Community Management
Don't worry, it's not about "The Delphi Community" but about this Praxis forum. Many Delphi developers and also Embarcadero are interested in Delphi and AI. RAD Studio Smart CodeInsight & LM Studio: A Local AI Alternative. -
Where? In the Main form? Created with "self" in Main form? The Autocreated with Application.CreateForm in *.dpr?
-
Polywick Studio - Delphi and C++ Builder developer.
Die Holländer replied to PolywickStudio's topic in Job Opportunities / Coder for Hire
polywickstudio.com -
See how after this session Sherlock could basically craft one long prompt based on the entire session that might let the AI output the right code in one time..
-
This statement suggests that AI knows Delphi in the same way a human programmer does by memorizing syntax, library names, and rules. But that's not how AI works. AI models don't think in terms of syntax or words. Instead, they process tokens which are small pieces of text (like parts of words, symbols, or code). During training, the model learns statistical patterns between these tokens. So, when you prompt the AI, it doesn't recall Delphi syntax. It generates output by predicting the most likely next token based on the prompt you gave and patterns it learned from training data. Just like AI doesn't understand English or Dutch, it doesn't understand Python or Delphi. It doesn't need to know the language it just needs to have seen enough examples to generate statistically likely and contextually correct output. In this way it can "inventing" non-existant functions. So, the quality of the output depends more on how much Delphi code the model has seen, how well the prompt guides the model and if the task is common or not. (Use translation if the task is more common in Python for example) Translation to Delphi might work well if the model has indeed seen enough Delphi examples and the prompt is clear. It doesn't rely on knowing syntax it relies on learned token patterns.
-
AI models don’t “pick up” algorithms in the way humans do. They learn statistical patterns in data,not abstract algorithmic concepts. (only when explicitly trained or fine-tuned to do so) They generate outputs based on learned correlations, not by executing abstract reasoning. So, They don't "just parroting of code it gleaned elsewhere". One thing many people overlook is how important the prompt is. It’s not like using a search engine. The way you ask the question can strongly affect the quality of the answer. For example, if you want Delphi code, a smart way to prompt is: Start with Python, which the model knows well: “Write a Python function that [clearly describe the task].” Then ask to translate it to Delphi, (yes, its a language model..) “Now translate this Python function into Delphi (Object Pascal), using Delphi syntax and conventions.” Add context if needed: “Assume the Delphi code will be used in a VCL application. Use appropriate types and structures.” This helps the model give better results, even for languages like Delphi that have less training data.
-
Migrating a Delphi XE4 System to MSSQL, PostgreSQL, and Linux – Softacom’s Experience
Die Holländer replied to Serge Pilko's topic in Tips / Blogs / Tutorials / Videos
"The (Linux) modernization allows the company to attract more clients." Can you explain more about the use of Linux the support? Are you talking about a GUI application? -
TGridify, convert a flat table into a pivot-grid in one line of code
Die Holländer replied to david berneda's topic in Delphi Third-Party
Yes, what happened with her between 2001 and 2003 ??- 3 replies
-
- pivot-table
- datamining
-
(and 2 more)
Tagged with:
-
Because as Remy suggested these 2 other units don't have DateUtils or SysUtils in the uses? But if you change the "DaysInMonth:=" into "Result:=" does that work ?
-
I didn't know that you can use a case statement where the numbers used are not in order.. 1,3,5,7,8,10,12:DaysInMonth := 31; And in the next line you use 4,6,9 and 11 while the line before ends with 12.. 4,6, 9,11:DaysInMonth := 30; What if you replace all the old Pascal style return vars DaysInMonth into Result like: 1,3,5,7,8,10,12:Result := 31;
-
Youtube comments - how are you retreaving them
Die Holländer replied to JohnLM's topic in Network, Cloud and Web
So, is your parsing "slow" or the fetching the API data from Youtube? How did you do the parsing? By running through the Memo of your screenshot? -
Youtube comments - how are you retreaving them
Die Holländer replied to JohnLM's topic in Network, Cloud and Web
Use the Youtube API. uses System.Net.HttpClient, System.JSON; procedure GetYouTubeComments(const VideoID, APIKey: string); var Client: TNetHTTPClient; Response: string; JSON: TJSONObject; begin Client := TNetHTTPClient.Create(nil); try Response := Client.Get( Format('https://www.googleapis.com/youtube/v3/commentThreads?part=snippet&videoId=%s&key=%s', [VideoID, APIKey]) ).ContentAsString; JSON := TJSONObject.ParseJSONValue(Response) as TJSONObject; // Parse JSON to extract comments finally Client.Free; end; end;