Anders Melander

So, can anyone remember which version of Delphi broke that scheme? It's not broken now, but I'm certain that there was a version after Delphi 6 that somehow broke {$IF} so it couldn't be used if backward and forward compatibility was needed.

For the same reason they have ridiculous NDAs and don't have public betas; They are so stuck in the 90s that they think anyone still gives a damn.n They never understood how to build a community.

Works for me:

Please stop using VERxxx. It's been possible to do {$if CompilerVersion >= XX} since Delphi 6 or 7. It's so annoying having to edit the .inc files of various 3rd party libraries with each new version of Delphi because they're designed as if the latest version = the last version. Graphics.32 uses {$LIBSUFFIX AUTO} for the D11+ packages. The earlier packages are manually maintained by using a diff tool to keep them synchronized with a master. The master usually comes from the latest supported Delphi version. There's no way I'm wasting disk space on old versions of Delphi just to maintain the package files. I have the latest version, and the few versions required by my job, and that's it.

Strange... It's like there's an echo in here.

Naming the constructor "Create" is just a convention. You can name it anything you like. Thus naming the constructor "From" will not make a difference. Here's your problem: Pascal isn't case sensitive so that code does nothing. Prefix your parameter names with "A" (i.e. name them AItem and AColor in this case) to avoid problems like that. FWIW, if you want to avoid memory allocation (i.e. allocate your object on the stack instead of the heap), make it a record instead.

There is no default 256 color palette. Old version of Windows "reserved" 16 fixed colors while new versions reserve 20 fixed colors. https://en.wikipedia.org/wiki/List_of_software_palettes#Microsoft_Windows_default_20-color_palette Changing the pixelformat to pf8bit forces the bitmap to use a color palette. I.e. instead of each pixel explicitly specifying the RGB color values, the pixels will contain an index into a palette of 256 colors. The colors in a palette generated by Windows will partially depend on the system palette which again depends on the Windows theme. https://learn.microsoft.com/en-us/windows/win32/gdi/system-palette-and-static-colors https://learn.microsoft.com/en-us/windows/win32/gdi/default-palette If you want a predictable result then you need to generate the palette using a predictable method: Either use a static palette or generate one using color quantization. There are methods in the GifImg unit for both solutions. https://docwiki.embarcadero.com/Libraries/Sydney/en/Vcl.Imaging.GIFImg.ReduceColors

Not necessarily. An emphasis on a specific library is most likely caused by existing use and thus an investment in that library. Regardless of "bestness".

Why is that interesting?

Support for FireDAC, one could hope

I can confirm that it doesn't work in NCSA Mosaic 2.0 I also tried Netscape Navigator 3.0

"Irrelevant" is the word you're looking for 🙂 If it works for you with 11.x then there is no problem.

Then it's probably a problem with your browser. [Ctrl]+F5 to do a forced reload or use another browser.

You need to have someone compile and run your project with 10.4.2 in order to conclude that. Right now you only know that your exe reproduces the problem and that an exe produced by one other person doesn't.

To be "Musked" I believe it's called

Strange. The URL seems correct so it's probably just a temporary hiccup with the redirection. 2: You translate the application using the component. If you use the component in multiple projects then you will have to do the translation for each project. That's unfortunately just the way Delphi localization is designed. You can save the translations to the Translation Memory and share that between projects, but it's an almost completely manual process. All the translations are contained in the translation project file (the xlat file). A translator will need amTranslationManager.exe and the .xlat file. I would recommend that you also create a portable-mode configuration file so the translator doesn't need to install anything. I think the docs mention that. Once you receive an updated xlat file from the translator you will have to merge that into your local xlat file using some sort of text merge tool. I suggest placing the file under version control and then use that to do the merge. There is no functionality in BTM to merge two translation project files, although that would be a handy feature. It's still there: https://bitbucket.org/anders_melander/better-translation-manager/src/master/#markdown-header-deploying-a-localized-application The hello_world demo contains an example of how to allow the user to choose the language. As far as I remember there are some other methods buried in the announcement thread: If your text is written as a string (or stringlist) property, then it can be translated. Yes. Individual properties on individual components are individually translated. BTM will help you with the translation of identical texts, but it will not do it automatically since that might not be what you want. Often context matters. For example, if you translate "Cancel"->"Fortryd" then it will prompt you to translate all other occurrences of "Cancel" using the same translation. It will even offer to translate texts that "almost" match such as "cancel", "&Cancel", "Cancel!", etc. Use resourcestrings just like you would use a string constant. The only thing to be aware of is that the resourcestring name needs to be unique within the unit. Resourcestrings are a language feature and as such it is processed by the compiler and the linker (with help from the RTL at runtime). FormCreate is a run-time thing so that doesn't matter; Use them where ever you want. Some prefer to keep all resourcestrings in a separate unit, some prefer to have them at the top of the unit and some to have them declared locally where they are used. I usually declare them with the most restrictive scope I can get away with: unit Foo; implementation procedure DoFoo; resourcestring sFooPublic = 'This is a public, global resourcestring'; // Usable by all that use unit Foo interface resourcestring sFooPrivate = 'This is a private, global resourcestring'; // Usable within unit Foo procedure DoFoo; resourcestring sFoo = 'This is a local resourcestring'; // Usable within DoFoo begin ShowMessage(sFoo); end; end. Note though that FreePascal doesn't support local resourcestrings. In FPC they must be declared globally (i.e. outside the methods/functions).

One needs to take into account that we're not dealing with lines but line segments. Lines have infinite length, while line segments have finite length. We don't want to detect a hit beyond the end of the line segment: I think the following one does the trick. I use it in a bitmap editor for scanline conversion of lines (the above image was made with it). (* Distance from point to line segment. Let A=(xa,ya), B=(xb,yb), X=(x,y) Now, the line between A and B is given parametrically by V(k) = (1-k)A + kB = A + k(B - A) and adding the constraint 0 <= k <= 1 makes V(k) the line segment from A to B. Now, the line through X perpendicular to V(k) intersects V(k) when k equals (B - A) . (X - A) k = ------------------- | B - A |^2 So if k <= 0, X is closest to A, if k >= 1, X is closest to B, and if 0 < k < kp, X is closest to V(k). *) function DistanceToLine(X,Y, XA,YA, XB,YB: integer; var RunLength: Single): Single; function VectorLength(X1,Y1, X2,Y2: Single): Single; inline; begin Result := Hypot(X1-X2, Y1-Y2); // = Sqrt(Sqr(X1-X2) + Sqr(Y1-Y2)) end; var k: Single; dx, dy: integer; begin dx := XB-XA; dy := YB-YA; if (dx <> 0) or (dy <> 0) then begin k := (dx*(X-XA) + dy*(Y-YA)) / (Sqr(dx)+Sqr(dy)); if (k <= 0) then // Point before or at start of line segment Result := VectorLength(XA,YA, X,Y) else if (k >= 1) then // Point after or at end of line segment Result := VectorLength(XB,YB, X,Y) else // Point within line segment Result := VectorLength(X,Y, XA+k*dx, YA+k*dy); // = VectorLength(X,Y, (1-kp)*XA+kp*XB, (1-kp)*YA+kp*YB); RunLength := k; end else // Degenerate line begin RunLength := 0; Result := VectorLength(XA,YA,X,Y); end; end; The result is the distance, the RunLength parameter can be used to determine where the projection of the point onto the line lies: before, on, or after the line segment. A small optimization can be done since we're doing hit-testing and don't really need the actual, precise distance value. The VectorLength function uses Hypot (i.e. Pythagoras) to calculate the distance. Hypot internally does a Sqrt, which is an expensive operation. The Sqrt can be eliminated so we return the squared distance instead and then we just need to compare that value against the squared hit limit instead. I.e. if your hit limit was 4 then compare the squared distance against 4*4 = 16 instead.

I have attached an extraction of the BTM documentation I have. As I said it was written for use with a specific project (called FooBar in this version) and as such contains references to the setup of that project (folders, config files, etc.) and it assumed that the initial translation project file and update have already been done. The source document, as mentioned, comes from Confluence so the formatting of the exported document is a bit wacky, but I think it should be readable. Using_Better_Translation_Manager_-_Anders.Melander_-_Confluence_(2023-08-16_22_10_10).html

I have no idea what you are referring to.

Ah, that... Um... Eh... I don't really have anything besides what's on the Bitbucket page. Well, actually, now that I think of it I think I have something in a corporate Confluence page somewhere, but I will have to revise it for public consumption and it's probably targeted for use with a specific commercial project. I'll look into it later tonight.

Done. The new version is 2.0.8628.33873

What help is that? Yes, I see it. Apparently, the feature branch wasn't merged correctly. I'll get that fixed.

Convert the bezier to a polyline with a fixed limited number of points (whatever drawing routine you end up using already does this internally, you just need a lot fewer points for hit testing than for drawing a smooth line). Iterate the segments of the polyline and find the minimum distance from your point to the segments. Even with thousands of beziers, you should be able to do this in no time at all (I'm guessing low milliseconds).

Fixed. The purge was rejected unless the project contained obsolete items with translations. Get the new release (2.0.8628.30767) here: https://bitbucket.org/anders_melander/better-translation-manager/downloads/ Changes since the last release (2.0.8370.39904): Fix for unable to purge project unless it contains Unused translations. #43: Specify external PE resource module stub via command line #41: DeepL Integration Added Save As... action. Previously the Save action always prompted for filename.

Reproduced. I'll investigate now.