Jump to content


  • Content Count

  • Joined

  • Last visited

  • Days Won


Everything posted by lynkfs

  1. lynkfs


    sounds like a duplication
  2. It becomes even more interesting when combining css variables and calc : since .... it has been possible to define css variables, which can be referenced in multiple ways they can even be set/get in pascal/js : get a handle to a stylesheet, or create one on the fly //stylesheet var style := browserapi.document.createElement("STYLE"); browserapi.document.head.appendChild(style); var styleSheet := style.sheet; set a css variable ("--variable-width") and insert into the stylesheet css variable names can be any string but have to start with -- //set initial css variable (width) var s1 := #' :root { --variable-width: 500px; }'; styleSheet.insertRule(s1, 0); retrieve and display css width variable browserapi.window.alert( browserapi.window.getComputedStyle(browserapi.document.documentElement) .getPropertyValue('--variable-width')); set css variable to some value browserapi.document.documentElement.style['--variable-width'] := Memo1.handle.getBoundingClientRect().width; use this variable in a 'calc'ulation Memo1.handle.style['height'] := 'calc(var(--variable-width) * 0.6)'; having a 'width' variable is just an example, maybe a bit contrived. Works fine though. Usually css variables are used for colors etc, and those values which need to be set globally for the whole application.
  3. lynkfs


    Off topic : there is an interesting effect when putting images on top of images. Take an image and give it a large width and height. Take the same image (same src) and place it on top of the first image, give it same height but say 25% of width and center it. Both images should have object-fit set to 'cover' and reduce opacity of the first one to about 65%. Depending on the image, about 1 in 5 give a passable interesting result doing this, 1 in 50 look amazing. Some examples here. (not optimised for size, so slow loading)
  4. Percentage based positioning has the advantage of having a 'responsive' page layout with little effort. Responsive in this respect means that page layouts will scale seamlessly on devices which may have slightly different pixel dimensions horizontally and/or vertically. So instead of having an editbox on absolute positioning 100,100,400,35 in pixel terms, it is sometimes easier to define all or some of these in percentages. Conveniently 10%-left is interpreted by the browser as starting on 10% of the viewport width, 10%-top starts at 10% of the viewport height etc. Mixing pixels and percentages works fine too MyEditBox.handle.style.left := '10%'; MyEditBox.handle.style.top := '12px'; //MyEditBox.left := 12; This is all still based on absolute positioning, which is the default in Smart. An alternative approach, with the same outcome, is to define dimensions at runtime based on a formula. MyEditBox.handle.style.left := 'calc(10vw - 20px)' resolves in a left positioning of 10% of the viewport-width minus 20 pixels : 28 pixels on a 480 pixels wide phone Note that the minus sign needs to be enclosed in spaces or this won't work This can be also used to scale f.i. font size depending on device capabilities : MyEditBox.handle.style['font-size'] := 'calc(16px + 1vw)' which has the effect of a font size which grows by 1 pixel for every 100 pixels of viewport width. It can get a bit funny on orientation change an all-percentage example page with some random text and images, based on these principles, looks like this , which scales nicely from mobile to desktop (resize browser window).
  5. Basically you'll want to add a history entry whenever you change forms. Also make up alias url's for every form. For instance 'Form1' can have an url-alias of 'index.html', 'Form2' an alias of 'bookreviews.html' etc. Every history entry links a Form (from) to an alias (to), see below. There is an old post on this forum (2015), but checking my last project I did this : Put an initial history entry on startup and also initiate the main eventhandler for handling backbutton presses unit Unit1; interface uses Pseudo.CreateForms, // auto-generated unit that creates forms during startup System.Types, SmartCL.System, SmartCL.Components, SmartCL.Forms, SmartCL.Application; type TApplication = class(TW3CustomApplication) procedure ApplicationStarting; reintroduce; end; implementation uses Globals; procedure TApplication.ApplicationStarting; begin browserapi.window.onpopstate := procedure(e:variant) begin CBOnPopState(e); end; browserapi.window.history.pushState(nil,'Form1','index.html'); end; end. Whenever there is a change in form, add an entry to the history object browserapi.window.history.pushState(nil,'Form1','bookreview.html'); //from Form1 to bookreview.html The second parameter (Form1) is the 'from' Smart form-name and the third parameter is the 'to' fake url. Supposing that Form2 in this example handles the bookreviews, then at some stage users will return to Form1, at which time another history entry 'Form2' --> 'index.html' will be added. The handler simply reacts to the history popstate events : Procedure CBOnPopState(e: variant); Begin if (browserapi.window.location.pathname = '/') then GoToForm('Form1'); if (browserapi.window.location.pathname = '/index.html') then GoToForm('Form1'); if (browserapi.window.location.pathname = '/bookreview.html') then GoToForm('Form2'); ... The above is for browser projects, but should work for native android as well
  6. I've used the history object in relation to the browsers backbutton many times. No problems on ios. Not your code exactly but similar
  7. lynkfs

    mouse position

    The 'featured demo' : Mouse Coordinates project compiles ok but results in error uncaught : Object not instantiated in TW3CustomControl._setMouseMove [line: 5783, column: 21, file: SmartCL.Components]
  8. lynkfs

    drag and drop

    I've spend hours to implement a simple drag & drop mechanism, and I've given up. The mechanism is not that difficult, but the browser implementations drive me nuts. The problem is that on different chrome browsers the 'drop' position is off from what it should be. If someone's interested, the code below works flawlessly in the internal ide browser. Executing on win10-latest chrome you'll see a jump in positioning of elements when dropped. I've finally figured out, that the jump-amount is based on where the element has been 'grabbed'. It is zero when grabbing top/left, and half the width of the grabbed element when grabbed bottom/right. Checking up on stack-overflow, there are heaps of references to this problem. Without a workable solution. And posts dating from 6 years ago until recently. One might expect these type of bugs to be resolved in a timespan that long. test-code procedure TForm1.InitializeForm; begin inherited; // this is a good place to initialize components Var Drag : JW3Panel := JW3Panel.Create(self); Drag.SetBounds(50,50,100,100); Drag.handle.style.border := '1px solid silver'; Drag.handle.setAttribute('draggable','true'); Var Drop : JW3Panel := JW3Panel.Create(self); Drop.SetBounds(50,300,400,200); Drop.handle.style.border := '1px solid silver'; Drag.handle.ondragstart := procedure(ev: variant) begin Drop.handle.ondragover := procedure(ev: variant) begin ev.preventDefault(); end; //payload = image.id, mouse-offsetX and mouse-offsetY ev.dataTransfer.setData("text", ev.target.id + ';' + inttostr(ev.offsetX) + ';' + inttostr(ev.offsetY)); ev.dataTransfer.effectAllowed := "copy"; ev.dataTransfer.dropEffect := "copy"; end; Drop.handle.ondrop := procedure(ev: variant) begin ev.preventDefault(); var data := ev.dataTransfer.getData("text"); //split payload into image.id, mouse-offsetX and mouse-offsetY var myarray := StrSplit(data,';'); var target : variant := document.getElementById(myarray[0]); ev.target.appendChild(target); target.style.left := ev.offsetX - StrToInt(MyArray[1]) + "px"; target.style.top := ev.offsetY - StrToInt(MyArray[2]) + "px"; end; end; I've ended up with a mouse-event based drag & drop solution rather than using the html5 based d&d api.
  9. lynkfs


    Digging in a bit further, see previous posts in this thread. I've set myself the task to define a workable styled css-only component : a slider. The structure is reasonably simple : in xml terms it will be <slider> <rail>...</rail> <knob>...</knob> </slider> The rail is an element which takes up the top half of the slider, with a bottom border color. The knob is an element with a border radius of 50%, which makes it a circle. The theme file for this component looks something like // slider theme01.slider.height: 100% // slider rail theme01.slider.rail.height: 50% theme01.slider.rail.border-bottom: 1px solid red // slider knob theme01.slider.knob.background-color: red theme01.slider.knob.height: 2em theme01.slider.knob.width: 2em theme01.slider.knob.left: 0% theme01.slider.knob.top: -1em theme01.slider.knob.border-radius: 50% This contains enough information to be able to construct and style a slider component on the fly. To make it a bit more interesting, I've added movement capabilities by adding a compiled procedure as a string : theme01.slider.knob.function01: onmousedown = function (e) { var saveX; saveX = e.clientX; Self.handle.onmousemove = function (e$1) { which can be just eval-ed at runtime. This function simply manipulates the 'left' property of the knob. So here we go, a working slider component from a theme file only. (the complete theme-file here). And since it does not have any absolute dimensioning info, it resizes automatically to its container.
  10. lynkfs


    Usually I try to steer away from external frameworks. They seldom provide exactly what is needed and are difficult to modify. An exception may be the ionic framework. It is an open source ui toolkit with some nice elements in it. The good thing is that their components subtly change depending on which platform they are used on. The following code uses 2 Ionic components : a button and an extremely elegant selector var iobutton : TIonicButton := TIonicButton.Create(self); iobutton.setbounds(100,100,100,35); iobutton.color := 'primary'; iobutton.fill := 'outline'; iobutton.caption := 'mybutton'; iobutton.onclick := procedure(sender:TObject) begin showmessage((sender as TIonicButton).caption); end; end); and the selector var iofab : TIonicFab := TIonicFab.Create(self); iofab.setbounds(400,200,56,56); iofab.mainIcon := 'share'; iofab.topIcon := 'logo-vimeo'; iofab.bottomIcon := 'logo-facebook'; iofab.leftIcon := 'logo-instagram'; iofab.rightIcon := 'logo-twitter'; iofab.right.button.onclick := lambda showmessage('twitter'); end; See demo here The code for both the button and the selector component in unit2/3 of this project
  11. lynkfs


    Looking into this idea a bit further. So the ultimate styler program needs to do this : - read in a theme file - discover the visual components in the theme - and populate them for use - have a canvas where these components can be dropped on - and can be manipulated qua size and position - and have a preview for how it looks in real life Basically this is an exercise in elevating css to its maximum capabilites, without using js (pascal) or html at all in the styling process. A first draft of such a program : The demo theme file contains components for buttons and images, and thus these are displayed on top. These can be dragged onto the canvas and positioned at will. The coordinates are displayed on the right, either in absolute or relative values. The preview icon on the canvas opens up a new live browser window. So far so good. The huge number of atomic css properties per visual component (280) now has been subdivided in 3 chunks : A- those css properties to do with visual design : x (left), y (top), width and height These are displayed on the right B- those css properties which are part of the theme file for this component so if for instance the theme file has a line like this: "theme01.buttons.generic.fontWeight: bold", then this will result in a modifier on the screen, where 'bold' can be changed. <to do> C- the remainder of the 280 css properties (280 minus A minus B). Those are either not important, have passable default values or are only seldomly important for styling of the particular component. <to do> Using components ultimately translates into html. Taking this example : <img class="img" id="Component38" src="images/buttons.jpg" name="button" style="visibility: visible; display: inline-block; position: absolute; overflow: auto; left: 350px; top: 10px; width: 82px; height: 34px; cursor: pointer;"> Styling through css results in modifying the style attribute (.. style="visibility: etc...) Still debating whether the other attributes should be part of the styling process. In the demo above for instance I've assigned the 'src' attribute of images to a random unsplash picture, instead of making it directly editable. Its not a css property after all. Demo here (alpha version)
  12. lynkfs

    Scroll form

    There is a timer control in the system.time unit memo1.Text := #'Apple Banana Cherry'; Memo1.Text := 'Apple' + #10 + 'Banana' + .. Memo1.Add('Apple',true); Memo1.Add('Banana',true); all give the same result
  13. lynkfs

    Scroll form

    I'm not a js guru , or maybe Anyway, logically it's either the form or the memo which reacts to scrolling/swiping. If you want both, you need to be able to distinguish in which situations you want either to happen. There are various ways to do this, one of them this : as soon as the form scrolls, disable scrolling on the memo altogether. Then you need to determine under which condition you want to re-enable scrolling the memo. One way would be to have a simple onclick handler procedure TForm1.InitializeForm; begin inherited; // this is a good place to initialize components self.NativeScrolling := true; self.onscroll := lambda W3Memo1.NativeScrolling := false; end; W3Memo1.OnClick := lambda W3Memo1.NativeScrolling := true; end; demo here Instead of having to click the memo, it would be better to do this automatically when form-scrolling has ended. There is no such thing as a scroll-end event, but you could simulate one by using a timer of say 1000 ms. Reset the timer continuously during form scrolling and re-enable memo scrolling after the timer fires. This may work for you, but it is hooking into some finely tuned browser actions, so look out for surprises.
  14. lynkfs


    Styling of visual elements. There are some 280 separate css properties which can be used to style each and every visual dom element. The following statement produces such a list var styles := browserapi.window.getComputedStyle(Panel.handle); (280 entries) If we disregard the properties which have a vendor prefix (-webkit, -moz etc) we still end up with some 230 properties for each visual element. And although many of these have default property values, the complexity is just too great to do styling on this level. Each and every css framework (including Smart) tries to reduce the complexity by a number of measures group css properties into meaningful categories, give these groups meaningful names and possibly also rename individual properties connect groups to visual components making sure that all groups adhere to application wide values (i.e. primaryBlue is set to "#ebf8ff" in all groups) eliminate annoying differences in how browsers interpret universal css do this either static and/or dynamic in code The way the above is implemented varies wildly across frameworks. Two examples : Smart does its standard styling mainly by grouping classes on the level of components (1), giving them component names like .TW3Button (2,4). Every visual component gets this class on creation with 2 additional classes for background and border (4). So the resulting styling of a button gets defined by <button class="TW3Button TW3ButtonBackground TW3ButtonBorder" ... The supplied css templates in the Templates directory have a feature which implements application wide setting (5), so f.i. const 'EdgeRounding' will be set to "4px" in all components which use 'EdgeRounding (3)'. There also some settings which standardise browser behaviour (6) Furthermore the rtl has methods to manipulate all classes, groups, properties and stylesheets in code (SmartCL.Css.Classes, SmartCL.Css.Stylesheet (7) Tailwind CCS. This framework is sort of on the other side of the spectrum. It groups css properties in a large numer of very small groups (1,2,3). These groups are almost atomic in nature, so f.i. the 'm-0' class is specified as { margin: 0 }. It basically does not do any of 4,5,6 or 7 So implementing a button using Tailwind involves many classes : <button class="bg-blue-500 hover:bg-blue-700 text-white font-bold py-2 px-4 rounded" which gives a light blue button with rounded corners, becomes a bit darker on hover, has its caption in bold white and has specific values for top/bottom and left/right padding. Nevertheless people build beautifully styled apps and websites using Tailwind. And it is readily accessible through code (7) Thinking about how to merge the best features of the above into a styling design mechanism, I would look at doing this : 1- define a (limited) number of atomic groups. An atomic group combines and names simple css properties which can be used in more complex structures. Candidates are : colors, borders, backgrounds and fonts to start off with. The font-category f.i could be sub-grouped in 'families', 'sizes', 'weights', 'letterSpacings' and 'lineHeights'. As an example the 'lineSpacings' subgroup could then be specified as "letterSpacings": { "tighter": "-0.05em", "tight": "-0.025em", "normal": "0", "wide": "0.025em", "wider": "0.05em", "widest": "0.1em" } 2- define generic building blocks or component groups, i.e a generic button, using as many of the atomic groupings from 1) as possible "button": { "cursor": "pointer", "fontSize": "100%", "lineHeight": "inherit", "backgroundColor": { "colors" : "primary" }, "border": "none", "color": "white", "fontWeight": "bold", "borderRadius": { "radii" : "default" }, ":hover": { "backgroundColor": { "colors" : "primaryHover"}, } This uses the atomic 'colors' group, which will have entries like "primary": "#2b6cb0", "primaryHover": "#2c5282" which then will be substituted in the above, so f.i. the 'backgroundColor' entry will become "backgroundColor": "#2b6cb0" This way all groups who use 'colors:primary' will have the same colour value, application-wide. 3- define descendants from the component groups from 2), f.i. a rounded pill button, which only has those properties different from its parent "buttons": { "pill": { "borderRadius": "full", } } So far the above examples boil down to something like this after substitution (2) and child creation (3) theme01.colors.primary: #2b6cb0 theme01.colors.primaryHover: #2c5282 theme01.buttons.generic.cursor: pointer theme01.buttons.generic.fontSize: 100% theme01.buttons.generic.lineHeight: inherit theme01.buttons.generic.backgroundColor: #2b6cb0 theme01.buttons.generic.border: none theme01.buttons.generic.color: white theme01.buttons.generic.fontWeight: bold theme01.buttons.generic.borderRadius: 0.25rem theme01.buttons.generic.hover.backgroundColor: #2c5282 theme01.buttons.generic.pill.borderRadius: 9999px and a class invocation like <button class="buttonPill ..." (or even TW3Button) 4- generate this into a regular css template and include the modernize.css entries to harmonize across browsers as well. 5- integrate this in the visual designer Actually the style generation should closely follow the hierarchy in the visual designer. So if there is a rounded-button on a panel on a form on the <body> element, then the css property values should reflect the #body, #form, #panel, #generic-button and #rounded-button properties (and nothing more). just an idea, but doable
  15. lynkfs


    Every now and then I read something so refreshing, that it jolts the old braincells out of its rut. 10/GUI describes a new gui system concept which does away with keyboards and mice, relying on gestures instead Fun, and somehow very convincing Recently I needed a mechanism to cycle through an image carousel. There are different ways of doing this, but for touch enabled devices I opted for a one-finger swipe gesture. W3Image.OnSwipeLeft := procedure(sender: TObject) begin ... end; The implementation of OnSwipeLeft below is somewhat project specific, but includes the following - it is made a property of all customcomponents descendants (helper construct) - it extends TNotifyEvent, so it does not only return the sending object (sender: TObject), but also the event object (event: variant). That way OnSwipeLeft has access to properties of both (and handles all event-types) - it checks if the device has touch capabilities and uses the touch-api if that is the case, otherwise it uses pointerevents Demo project (covering all swipe orientations. Verbose code, can be simplified) The touch-api supports multi-touch events to a certain maximum. My phone can handle 5 fingers. That makes it possible to implement for instance two-finger swipes to slide-in the next form. Form.OnFormSwipeRight := lambda Application.GotoForm('Form2', feFromRight); end;
  16. you probably can use the same mechanism W3StringGrid1.handle.GetChildById(W3StringGrid1.tagid + '_row_' + inttostr(i) + '_col_' + inttostr(4)).style['color'] := 'red';
  17. not the most straightforward way of doing this, but this works : procedure TForm1.W3StringGrid1CellClick(const Sender: TObject; const Row: Integer; const Col: Integer); begin //var elem := browserapi.document.getElementById(W3StringGrid1.Handle.id + '_row_' + inttostr(row) + '_col_' + inttostr(col)); //either var elem := W3StringGrid1.handle.GetChildById(W3StringGrid1.tagid + '_row_' + inttostr(row) + '_col_' + inttostr(col)); //or elem.style['color'] := 'red'; end; maybe a handle to a rendered cell could be made a standard property
  18. lynkfs


    Edited Had a look at vue, but never used it. After some more testing of the proxy object, my conclusion is that you are right. Proxy objects track 'reads' across all properties of the tracking object, any level deep. 'Writes' though are only tracked on the same level as where they are specified. Sort of logical as a set-handler only gets as input an object, property and value, and as properties can have the same name across levels, it wouldn't know what to track. It is feasible though to set subsequent proxies on deeper levels, and code specific set-handlers for it ...
  19. works as expected on my system is the error the same when you navigate to ..\output\ and start node there ? (node WebSocketServer.js)
  20. lynkfs


    Correct. 😃
  21. just as a proof of concept : tutor initiates conversation with student : https://lynkfs.com/Experiments/peer/receive/www and on another machine : https://lynkfs.com/Experiments/peer/send/www (must be on another machine) student sends a text by copying tutor id into the edit box and pressing buttons 1, 2 and 3 this text then arrives at the tutors project The user interface of these projects is really bad, sorry, just proof-of-concept. This demo is basically a simple chap app, but since it is using webrtc (and webrtc handles not only text but also other media), I feel that extending this to a full-fledged channel between parties should be feasible and reasonably straightforward
  22. Just going through the first bit of their docs, it looks fairly straightforward procedure TForm1.W3Button1Click(Sender: TObject); begin // var Script := browserapi.document.createElement('script'); Script.src := 'https://cdn.jsdelivr.net/npm/peerjs@0.3.20/dist/peer.min.js'; browserapi.document.head.appendChild(Script); Script.onload := procedure begin writeln('peer.js loaded'); asm @mypeer = new Peer({key: 'lwjd5qra8257b9'}); end; //demo key mypeer.on('open', procedure(id:string) begin writeln('New peer ID : ' + id); end); end; end; the remainder doesn't look too cumbersome, i'll have a look later
  23. lynkfs


    The proxy object does handle deeper levels as well code below refers to earlier post : //set proxies var MyProxy : variant := new JObject; var ProxyObj : variant := new JObject; var leveltest: variant := new JObject; ProxyObj.leveltest := leveltest; ProxyObj.leveltest.level := 'initiallevel'; Button2.OnClick := lambda Button2.Caption:= MyProxy.leveltest.level; end; var handler : variant := new JObject; handler.get := function(obj, prop: variant): variant begin writeln(json.stringify(obj)); Memo1.Add('someone read ' + prop + ' as ' + obj[prop].level + #10); result := obj[prop]; end;
  24. webrtc I think has come of age, so it's worthwhile exploring I'll be happy to have a look
  • Create New...