The following is verbatim of a feedback report (FB19809442) I submitted, shared here as someone else might be interested to see it (I hate the fact that we can't see each other's feedbacks).
On iOS 16, TextKit 2 calls NSTextLayoutFragment's draw(at:in:) method once for the first paragraph, but for every other paragraph, it calls it continuously on every scroll step in the UITextView. (The first paragraph is not cached; its draw is called again when it is about to be displayed again, but then it is again called only once per its lifecycle.)
On iOS 17, the behavior is similar; the draw method gets called once for the 1st and 2nd paragraph, and for every other paragraph it again gets called continuously as a user scrolls a UITextView.
On iOS 18 (and iOS 26 beta 4), TextKit 2 calls the layout fragment's draw(at:in:) on every scroll step in the UITextView, for all paragraphs. This results in terrible performance.
TextKit 2 is promised to bring many performance benefits by utilizing the viewport - a new concept that represents the visible area of a text view, along with a small overscroll. However, having the draw method being constantly called almost negates all the performance benefits that viewport brings. Imagine what could happen if someone needs to add just a bit of logic to that draw method. FPS drops significantly and UX is terribly degraded.
I tried optimizing this by only rendering those text line fragments which are in the viewport, by using NSTextViewportLayoutController.viewportBounds and converting NSTextLineFragment.typographicBounds to the viewport-relative coordinate space (i.e. the coordinate space of the UITextView itself). However, this patch only works on iOS 18 where the draw method is called too many times, as the viewport changes. (I may have some other problems in my implementation, but I gave up on improving those, as this can't work reliably on all OS versions since the underlying framework isn't calling the method consistently.)
Is this expected? What are our options for improving performance in these areas?
TextKit
RSS for tagManage text storage and perform custom layout of text-based content in your app's views using TextKit.
Posts under TextKit tag
47 Posts
Selecting any option will automatically load the page
Post
Replies
Boosts
Views
Activity
The NSTextViewportLayoutControllerDelegate.textViewportLayoutControllerDidLayout(_:) documentation states that
Layout information on textViewportLayoutController is up-to-date at the point of this call.
however it is easy to put the NSTextViewportLayoutController in a state where after calling textViewportLayoutControllerDidLayout, the value of viewportRange is nil (unexpected) and value of the property viewportBounds is .zero
The TextKit2 sample application found at https://developer.apple.com/documentation/uikit/using-textkit-2-to-interact-with-text makes that assumption as well, and in few places force unwrap the value of viewportRange, that leads to runtime crashes.
This behavior is also discussed in Developer Forum thread about TextKit2 viewport relocation: https://developer.apple.com/forums/thread/761364?answerId=800516022#800516022
How to reproduce:
Run Mac target of LayoutWithTextKit2 sample project found at https://developer.apple.com/documentation/uikit/using-textkit-2-to-interact-with-text
locate menu.rtf file and duplicate its content several times - the goal is to increase the length of the layout text
quickly resize application window - that results in viewport layouts - that result in out-of-bound viewport - that results ina crash
OR quickly scroll down/up to the end of the document using scroller bar on the right side of the window
Reproducible 100%
The situation occurs when the document is not fully laid out, the estimated size (height) of the content exceeds the final (correct) height, and the layoutViewport() function is executed quickly. Resulting in partial viewport layout, and once the viewport moves outside of the document's total height, the viewportLayoutController starts to report viewportRange = nil.
FB19698121
Why does it happen? Is it expected? How to recover from that state? And most importantly, how to make the NSTextLayoutManager display the portion of the document that is currently scrolled to. and how to do it without for ce layout the full document on each viewportLayout()
How can we performantly scroll to a target location using TextKit 2?
Hi everyone,
I'm building a custom text editor using TextKit 2 and would like to scroll to a target location efficiently. For instance, I would like to move to the end of a document seamlessly, similar to how users can do in standard text editors by using CMD + Down.
Background:
NSTextView and TextEdit on macOS can navigate to the end of large documents in milliseconds. However, after reading the documentation and experimenting with various ideas using TextKit 2's APIs, it's not clear how third-party developers are supposed to achieve this.
My Code:
Here's the code I use to move the selection to the end of the document and scroll the viewport to reveal the selection.
override func moveToEndOfDocument(_ sender: Any?) {
textLayoutManager.ensureLayout(for: textLayoutManager.documentRange)
let targetLocation = textLayoutManager.documentRange.endLocation
let beforeTargetLocation = textLayoutManager.location(targetLocation, offsetBy: -1)!
textLayoutManager.textViewportLayoutController.layoutViewport()
guard let textLayoutFragment = textLayoutManager.textLayoutFragment(for: beforeTargetLocation) else {
return
}
guard let textLineFragment = textLayoutFragment.textLineFragment(for: targetLocation, isUpstreamAffinity: true) else {
return
}
let lineFrame = textLayoutFragment.layoutFragmentFrame
let lineFragmentFrame = textLineFragment.typographicBounds.offsetBy(dx: 0, dy: lineFrame.minY)
scrollToVisible(lineFragmentFrame)
}
While this code works as intended, it is very inefficient because ensureLayout(_:) is incredibly expensive and can take seconds for large documents.
Issues Encountered:
In my attempts, I have come across the following two issues.
Estimated Frames: The frames of NSTextLayoutFragment and NSTextLineFragment are approximate and not precise enough for scrolling unless the text layout fragment has been fully laid out.
Laying out all text is expensive: The frames become accurate once NSTextLayoutManager's ensureLayout(for:) method has been called with a range covering the entire document. However, ensureLayout(for:) is resource-intensive and can take seconds for large documents. NSTextView, on the other hand, accomplishes the same scrolling to the end of a document in milliseconds.
I've tried using NSTextViewportLayoutController's relocateViewport(to:) without success. It's unclear to me whether this function is intended for a use case like mine. If it is, I would appreciate some guidance on its proper usage.
Configuration:
I'm testing on macOS Sonoma 14.5 (23F79), Swift (AppKit), Xcode 15.4 (15F31d).
I'm working on a multi-platform project written in AppKit and UIKit, so I'm looking for either a single solution that works in both AppKit and UIKit or two solutions, one for each UI framework.
Question:
How can third-party developers scroll to a target location, specifically the end of a document, performantly using TextKit 2?
Steps to Reproduce:
The issue can be reproduced using the example project (download from link below) by following these steps:
Open the example project.
Run the example app on a Mac. The example app shows an uneditable text view in a scroll view. The text view displays a long text.
Press the "Move to End of Document" toolbar item.
Notice that the text view has scrolled to the bottom, but this took several seconds (~3 seconds on my MacBook Pro 16-inch, 2021). The duration will be shown in Xcode's log.
You can open the ExampleTextView.swift file and find the implementation of moveToEndOfDocument(_:). Comment out line 84 where the ensureLayout(_:) is called, rerun the app, and then select "Move to End of Document" again. This time, you will notice that the text view moves fast but does not end up at the bottom of the document.
You may also open the large-file.json in the project, the same file that the example app displays, in TextEdit, and press CMD+Down to move to the end of the document. Notice that TextEdit does this in mere milliseconds.
Example Project:
The example project is located on GitHub:
https://github.com/simonbs/apple-developer-forums/tree/main/how-can-we-performantly-scroll-to-a-target-location-using-textkit-2
Any advice or guidance on how to achieve this with TextKit 2 would be greatly appreciated.
Thanks in advance!
Best regards,
Simon
Hello 👋
I'm playing with the Apple TextKit2 sample app (particularly on macOS).
I found that on some long document the evaluated height given by enumerateTextLayoutFragments API is wrong (or at least not I expect) which imply I can no longer scroll in my document even if I have not reached the end of it (which is not what I expect as you can guess). I'm clearly missing a point here. I can reproduce it on the Apple sample app by only changing the text content:
https://developer.apple.com/documentation/UIKit/using-textkit-2-to-interact-with-text
Using TextKit2, is it my responsability as developer to check that I've reached end of the scrollview whether not being at the end location of the document and call some specific TextKit2 API to invalidate estimation or something ?
Here is an updated version of the Apple sample app with another text content that show the issue.
https://drive.google.com/file/d/1jtTD84oqGAG4_A9DfqFl_yHmbLKhF1e8/view?usp=sharing
Environment: Xcode 16.4 - macOS 15.6
If someone could help me with this, I would be extremely grateful. It puzzles me.
NB: I've observed that resizing the window a bit seems to force a new layout and make TextKit2 returns a more accurate height, ... until you reach the end of the document.
Hello!
I am trying to create an iOS app that is based around a very large, vertically scrolling text view. The text is broken up into many sections, and the user should be able to press buttons in the navigation, which programmatically scroll to those sections. The user can also change the font size in a settings menu. It should generally keep the user's spot when resizing fonts or rotating the screen (from portrait to landscape).
The problem I've been having is that no method of lazy text loading allows accurate enough navigation, and the text is too long to calculate the whole UI all at once. Here's my process in trying to find a solution:
My app is built in SwiftUI, so I started with a ScrollView and a LazyVStack, and I used .scrollPosition() and bound it to an Int?. It worked pretty well for most scroll locations both on screen and far off the screen, but when I programmatically scroll to a location that is off the screen but not very far off, it completely misses.
So, I investigated UIKit, and found that UITextView was a much better fit for the way I wanted to present the long text. I could also programmatically navigate by storing the NSRange of each section.
I tried to use scrollRangeToVisible(), but for long distance it would scroll so that the desired section was just below the viewport and thus off screen. Then I tried to use UITextView's textLayoutManager.textViewportLayoutController.relocateViewport() to send it to the correct NSTextRange, it would not jump all the way, but instead would do nothing until I tried to scroll again and it would jump slightly forward. I tried to use textViewportLayoutController.layoutViewport() after the jump, and that fixed the glitch when scrolling, but it still did not jump to the correct place, only slightly forward.
Then, I looked into TextKit 2 and the way it worked to try to find a solution. From what I can tell, it seems that to affect the NSTextViewportLayoutController without having to rewrite it, I need an NSTextViewportLayoutControllerDelegate, but the delegate required me to manually lay out the views, and in all the examples I've seen that use a custom NSTextViewportLayoutControllerDelegate, they wrote their own custom text view instead of using the default UITextView.
I started looking into writing a custom text view so I can get the programmatic scroll to work consistently. However, it felt like, from a maintainability standpoint, it would probably be best to stick with what Apple has already implemented.
For now, I found a workaround that scrolls consistently. Here is the code:
if let start = self.textView.position(from: self.textView.beginningOfDocument, offset: desiredLineRange.location) {
let location = textView.caretRect(for: start)
self.textView.setContentOffset(CGPoint(x: 0, y: location.origin.y), animated: false)
}
if let start = self.textView.position(from: self.textView.beginningOfDocument, offset: desiredLineRange.location) {
let location = textView.caretRect(for: start)
self.textView.setContentOffset(CGPoint(x: 0, y: location.origin.y), animated: false)
}
It does the job, because the first time it gets close enough, and the second time it gets to the precise location, but it just feels like a bit of a hack to run the same code twice. I was wondering if anyone knows what I could be doing wrong and if Apple provides any solutions for this?
(Also, all my UIKit navigation attempts ran inside an @objc func, which I passed using button.addTarget(self, action: #selector(buttonTapped), for: .touchUpInside). Just so you know in case it may be a problem with the way Swift and UIKit handle concurrency/parallel tasks).
Thank you!
The app puts button values into a text view area and controls the cursor. Upon an IOS touch screen, cut or paste, the IOS cursor loses sync with the app cursor causing an address out of bounds and fails.
The IOS cut or paste changes the cursor position and the text.endIndex address by shrinking or expanding the text field. IOS doesn't know about the app cursorPosition. If IOS could update the app cursor position and the text.endIndex position would solve the problem. Or if the app knew about the IOS change could update the app cursor position and the text.endIndex.
The current work around is the user sets the cursor to the text.startIndex using an app navigating button before the touch screen. The app does not fail. The user then navigates the cursor using arrow buttons to another position.
To see this happen download the free SummaGramIPAD Trial.(13")
TSIs have been requested with engineer suggestions for well over one year. I hope someone could figure this out.
I just finished SummaGram iPhone and would like to host it.
Thanks for your help.
Charlie
hi everyone, any thought on how to implement a RichText(markdown) live editor using the new TextEditor(text: AttributedString, selection: AttributedTextSelection)?
Having issues like:
how to get the location of the selected text relatived to the TextEditor bounds, which is used to show and position a floating toolbar.
how to detect the cursor point is a new line and the content is "# ", (of couse now the user enter a space after the "#"), and if so, how to apply a H1 heading format to that line
much more issues like this, but these two are how to get me started, thanks
struct TextEditor8: View {
@State var text: AttributedString
@State var selection = AttributedTextSelection()
@State var isShowFloatingToolbar = false
@State var toolbarPosition: CGPoint = .zero
init() {
var text = AttributedString("Hello ✋🏻,Who is ready for Cooking?")
let range = text.characters.indices(where: \.isUppercase)
text[range].foregroundColor = .blue
_text = State(initialValue: text)
}
var body: some View {
ZStack(alignment: .topLeading) {
GeometryReader { geo in
TextEditor(text: $text, selection: $selection)
.onChange(of: selection, perform: { newV in
let v = text[newV]
if v.characters.isEmpty {
isShowFloatingToolbar = false
} else {
isShowFloatingToolbar = true
toolbarPosition = CGPoint(x: 20, y: 20) // how to get CGPoint relatived to TextEditor from selection
}
print("vvvv", v.characters.isEmpty)
})
}
if isShowFloatingToolbar {
FloatingToolbarAtSelection()
.position(toolbarPosition)
}
}
}
}
I noticed that sometimes TextKit2 decides to crop some text instead of soft-wrapping it to the next line.
This can be reproduced by running the code below, then resizing the window by dragging the right margin to the right until you see the text with green background (starting with “file0”) at the end of the first line.
If you now slowly move the window margin back to the left, you’ll see that for some time that green “file0” text is cropped and so is the end of the text with red background, until at some point it is soft-wrapped on the second line.
I just created FB18289242. Is there a workaround?
class ViewController: NSViewController {
override func loadView() {
let textView = NSTextView(frame: CGRect(x: 0, y: 0, width: 400, height: 400))
let string = NSMutableAttributedString(string: "file0\t143548282\t1970-01-01T00:00:00Z\t1\t1f40fc92da241694750979ee6cf582f2d5d7d28e18335de05abc54d0560e0f5302860c652bf08d560252aa5e74210546f369fbbbce8c12cfc7957b2652fe9a75", attributes: [.foregroundColor: NSColor.labelColor, .backgroundColor: NSColor.red.withAlphaComponent(0.2)])
string.append(NSAttributedString(string: "file0\t143548290\t1970-01-01T00:05:00Z\t 2\t0f6460d0ed7825fed6bda0f4d9c14942d88edc7ff236479212e69f081815e6f1742c272753b77cc6437f06ef93a46271c6ff9513c68945075212434080e60c82", attributes: [.foregroundColor: NSColor.labelColor, .backgroundColor: NSColor.green.withAlphaComponent(0.2)]))
textView.textContentStorage!.textStorage!.setAttributedString(string)
textView.autoresizingMask = [.width, .height]
let scrollView = NSScrollView(frame: CGRect(x: 0, y: 0, width: 400, height: 400))
scrollView.documentView = textView
scrollView.hasVerticalScroller = true
scrollView.translatesAutoresizingMaskIntoConstraints = false
view = scrollView
}
}
In order to create a UITextView like that of the Messages app whose height grows to fits its contents (number of lines), I subclassed UITextView and customized the intrinsicContentSize like so:
override var intrinsicContentSize: CGSize {
var size = super.intrinsicContentSize
if size.height == UIView.noIntrinsicMetric {
layoutManager.glyphRange(for: textContainer)
size.height = layoutManager.usedRect(for: textContainer).height + textContainerInset.top + textContainerInset.bottom
}
return size
}
As noted at WWDC, accessing layoutManager will force TextKit 1, we should instead use textLayoutManager. How can this code be migrated to support TextKit 2?
I have a UITextView that contains paragraphs with text bullet lists (via NSTextList). I also implement NSTextContentStorageDelegate.textContentStorage(_:, textParagraphWith:) in order to apply some custom attributes to the text without affecting the underlying attributed text. My implementation returns a new NSParagraph that modifies the foreground color of the text. I based this on the example in the WWDC 21 session "Meet Text Kit 2".
UITextView stops rendering the bullets when I implement the delegate function and return a custom paragraph. Why?
func textContentStorage(_ textContentStorage: NSTextContentStorage, textParagraphWith range: NSRange) -> NSTextParagraph? {
guard let originalText = textContentStorage.textStorage?.attributedSubstring(from: range) else { return nil }
let updatedText = NSMutableAttributedString(attributedString: originalText)
updatedText.addAttribute(.foregroundColor, value: UIColor.green, range: NSRange(location: 0, length: updatedText.length))
let paragraph = NSTextParagraph(attributedString: updatedText)
// Verify that the text still contains NSTextList
if let paragraphStyle = paragraph.attributedString.attribute(.paragraphStyle, at: 0, effectiveRange: nil) as? NSParagraphStyle {
assert(!paragraphStyle.textLists.isEmpty)
} else {
assertionFailure("Paragraph has lost its text lists")
}
return paragraph
}
I am trying to implement the NSTextViewDelegate function textViewDidChangeSelection(_ notification: Notification). My text view's delegate is the Coordinator of my NSViewRepresentable. I've found that this delegate function never fires, but any other delegate function that I implement, as long as it doesn't take a Notification as an argument, does fire (e.g., textView(:willChangeSelectionFromCharacterRange:toCharacterRange:), fires and is called on the delegate exactly when it should be).
For context, I've verified all of the below:
textView.isSelectable = true
textView.isEditable = true
textView.delegate === my coordinator
I can call textViewDidChangeSelection(:) directly on the delegate without issue.
I can select and edit text without issues. I.e., the selections are being set correctly. But the delegate method is never called when they are.
I am able to add the intended delegate as an observer for the selector textViewDidChangeSelection via NotificationCenter. If I do this, the function executes when it should, but fires for every text view in my view hierarchy, which can number in the hundreds. I'm using an NSLayoutManager, so I figure this should only fire once. I've added a check within my code:
func textViewDidChangeSelection(_ notification: Notification) {
guard let textView = notification.object as? NSTextView,
textView === layoutManager.firstTextView else { return }
// Any code I want to execute...
}
But the above guard check lets through every notification, so, no matter what, my closure executes hundreds of times if I have hundreds of text views, all of them being sent by textView === layoutManager.firstTextView, but once for each and every text view managed by that layoutManager.
Does anyone know why this method isn't ever called on the delegate, while seemingly all other delegate methods are? I could go the NotificationCenter route, but I'd love to know why this won't execute as a delegate method when documentation says that it should, and I don't want to have to implement a counter to make sure my code only executes once per selection update. And for more reasons than that, implementing via delegate method is preferable to using notifications for my use case.
Thanks for any help!
Hi, I've got an app that displays markdown in UITextView / NSTextView.
I would like it to behave like Notes app does, that is if user types the line start modifier, e.g:
"# "
or
"> "
I would like the keyboard to show a capitalized letters.
I've tried looking into overriding insertText - and it breaks the predictive text (can not insert white space).
I've tried implementing UITextInputTokenizer but no luck either.
Like I said, I see the Notes app does it perfectly, so maybe I'm missing something obvious or there is a way to do it without interrupting the auto-correction and predictive text.
Ideally same thing can be applied to the NSTextView as the app support both platforms.
I've posted a couple times now about major issues I'm having with NSLayoutManager and have written to Apple for code-level support, but no one at Apple has responded to me in more than two weeks. So I'm turning to the community again for any help whatsoever.
I'm fairly certain it's a real bug in TextKit. If I'm right about that, I'd love for anyone at Apple to take an interest. And better yet, if I'm wrong (and I hope I am), I'd be incredibly grateful to anyone who can point out where my mistake lies! I've been stuck with this bug for weeks on end.
The crux of the issue is that I'm getting what seemed to be totally incompatible results from back to back calls to textContainer(forGlyphAt:effectiveRange:) and lineFragmentRect(forGlyphAt:effectiveRange:withoutAdditionalLayout:)... I'd lay out my text into a fairly tall container of standard page width and then query the layout manager for the text container and line fragment rect for a particular glyph (a glyph that happens to fall after many newlines). Impossibly, the layout manager would report that that glyph was in said very tall container, but that the maxY of its lineFragmentRect was only at 14 points (my NSTextView's isFlipped is true, so that's 14 points measuring from the top down).
After investigating, it appears that what is happening under the hood is NSLayoutManager is for some reason laying out text back into the first container in my series of containers, rather than overflowing it into the next container(s) and/or giving me a nil result for textContainer(forGlyphAt:...)
I've created a totally stripped down version of my project that recreates this issue reliably and I'm hoping literally anyone at Apple will respond to me. In order to recreate the bug, I've had to build a very specific set of preview data - namely some NSTextStorage content and a unique set of NSTextViews / NSTextContainers.
Because of the unique and particular setup required to recreate this bug, the code is too much to paste here (my preview data definition is a little unwieldy but the code that actually processes/parses it is not).
I can share the project if anyone is able and willing to look into this with me. It seems I'm not able to share a .zip of the project folder here but am happy to email or share a dropbox link.
I'm building a SwiftUI app with a UITextView subclass, and it seems that the software keyboard doesn't trigger the pressesBegan or pressesEnded functions of UITextView. With a hardware keyboard, pressesBegan works as expected, allowing us to intercept key presses in our subclass.
I can't find any documentation about this, or any other forum posts (here or on Stack Overflow) that talk about a discrepancy between software and hardware keyboard behaviors, and I can't believe this is an intended behavior. Our app is a SwiftUI app, in case that's relevant.
Does anyone have any guidance? Is this a bug or am I not understanding this API? Any information or work arounds would be greatly appreciated.
I've made a sample project that demonstrates this issue, which you can grab from GitHub at https://github.com/nyousefi/KeyPressSample. To see this in action, run the sample project and start pressing keys. The hardware keyboard will print the key press at the top of the screen (above the text view), while the software keyboard won't.
In summation: I have a nasty bug where my layout manager is laying out text visually overlapping on top of other text, i.e., into a container that it should have left in the rear view as it continues to lay out into ensuing containers. Details below...
I'm coding a word processing app with some custom pagination that involves multiple pages, within which there can be multiple NSTextView/NSTextContainer pairs that represent single column or dual column runs of text.
I generate pagination data by using a measuring NSLayoutManager. This process ensures that no containers overlap, and that they are sized correctly for their associated ranges of text (i.e., non-overlapping, continuous ranges from a single NSTextStorage).
I determine frame sizes by a series of checks, most importantly, by finding the last glyph in a column. Prior to the code below, remainingColumnRange represents the remaining range of my textStorage that is of a consistent column type (i.e., single, left column, or right column). My measuring passes consist of my measuringLayoutManager laying out text into its textContainers, the final of which is an extra overflowContainer (i.e., == measuringLayoutManager.textContainers.last!) which I only use to find the last glyph in the second to last container (measuringContainer, which is thus == measuringLayoutManager.textContainers[count - 2])
let glyphRangeOfLastColumnChar = measuringLayoutManager.glyphRange(forCharacterRange: remainingColumnRange, actualCharacterRange: nil)
let lastGlyphIndex = NSMaxRange(glyphRangeOfLastColumnChar) - 1
measuringLayoutManager.ensureLayout(for: measuringContainer) // Not sure if this is necessary, but I've added it to insure I'm getting accurate measurements.
if measuringLayoutManager.textContainer(forGlyphAt: lastGlyphOfColumnIndex, effectiveRange: &actualGlyphRangeInContainer) == overflowContainer {
actualCharRangeInContainer = measuringLayoutManager.characterRange(forGlyphRange: actualGlyphRangeInContainer, actualGlyphRange: nil)
let overflowLoc = actualCharRangeInContainer.location
remainingColumnRange = NSRange(location: overflowLoc, length: remainingColumnRange.length - overflowLoc)
currentPage += 1
} else {
lineFragmentRectForLastChar = measuringLayoutManager.lineFragmentRect(forGlyphAt: lastGlyphIndex, effectiveRange: nil)
// Resize measuring container if needed.
let usedHeight = lineFragmentRectForLastChar.maxY
if usedHeight < measuringContainer.size.height {
measuringContainer.size = CGSize(width: measuringContainer.size.width, height: usedHeight)
} else if usedHeight == measuringContainer.size.height {
currentPage += 1 // we perfectly filled the page
} else {
// This would be an error case, because all cases should have been handled prior to arriving here. I throw an error. I have never fallen through here.
throw MyClass.anError
}
}
// I use the above data to create a PageLayoutItem, which is a struct that has frame data (CGRect/x,y,w,h), a containerIndex (Int), pageNumber (Int), textRange (NSRange), columnType (custom enum).
// After this I remove the overflowContainer, and continue to iterate through. This is inefficient but I'm simplifying my code to identify the root issue.
I don't explicitly use these containers when done with my pagination process. Rather, I use the PageLayoutItems I have created to generate/resize/remove textContainers/textViews for the UI as needed. My UI-interfacing/generating NSLayoutManager, which is of course assigned to the same NSTextStorage as the measuring layout manager, then iterates through my paginator model class' pageLayoutItems array to generate/resize/remove.
I have verified my pagination data. None of my frames overlap. They are sized exactly the same as they should be per my measurement passes. The number of containers/views needed is correct.
But here's the issue:
My views render the text that SHOULD appear in my final textContainer/textView as visually overlapping the text in my second to last textContainer/textView. I see a garble of text.
When I iterate through my UI textContainers, I get this debug print:
TextContainer 0 glyphRange: {0, 172}
TextContainer 1 glyphRange: {172, 55}
TextContainer 2 glyphRange: {227, 100} // this is wrong, final 31 chars should be in container 3
TextContainer 3 glyphRange: {327, 0} // empty range here, odd
I have tried setting textContainers for glyph ranges explicitly, via:
// Variable names just for clarity here
layoutManager.setTextContainer(correctTextView.textContainer!, forGlyphRange: correctGlyphRangeForThisContainer)
Debug prints show that I'm setting the right ranges there. But they don't retain.
I have tried resizing my final text container to be much larger in case that was the issue. No dice. My final range of text/glyphs still lays out in the wrong container and overlaps the other content laid out there.
Any help here?? I've scoured the forums and have been dealing with this bug for two weeks straight with no hope in sight.
tl;dr: UITextView does not auto layout when isScrollEnabled = false
I have a screen with multiple UITextViews on it, contained within a ScrollView. For each textview, I calculate the height needed to display the entire content in SwiftUI and set it using the .frame(width:height:) modifier.
The UITextView will respect the size passed in and layout within the container, but since UITextView is embedded within a UIScrollView, when a user attempts to scroll on the page, often they will scroll within a UITextView block rather than the page. They currently need to scroll along the margins outside of the textview to get the proper behavior.
Since I am already calculating the height to display the text, I don't want the UITextView to scroll. However, when I set isScrollEnabled = false, the text view displays in a single long line that gets truncated. I have tried
Setting various frame/size attributes but that seems to have zero affect on the layout.
Embedding the textView within a UIView, and then sizing the container, but then the textView does not display at all.
Setting a fixed size textContainer in the layoutManager but did not work.
There's a lot of code so I can't copy/paste it all, but generally, it looks like
struct SwiftUITextEditor: View {
@State var text: AttributedString = ""
var body: some View {
ZStack {
MyTextViewRepresentable(text: $text)
}
.dynamicallySized(from: $text)
}
}
struct MyTextViewRepresentable: UIViewRepresentable {
@Binding var text: AttributedString
let textView = UITextView(usingTextLayoutManager: true)
func makeUIView(context: Context) -> UITextView {
textView.attributedText = text
textView.isScrollEnabled = false
}
...
}
TLDR: NSLayoutManager's textContainer(forGlyphAt:effectiveRange:) and lineFragmentRect(forGlyphRange:effectiveRange:) are returning inconsistent results.
Context: I'm developing a word processing app that paginates from an NSTextStorage using NSLayoutManager. My app uses a text attribute (.columnType) to paginate sub-ranges of the text at a time, ensuring that each columnRange gets a container (or series of containers across page breaks) to fit. This is to support both multi-column and standard full-page-width content.
After any user edit, I update pagination data in my Paginator model class. I calcuate frames/sizes for the views/containers, along with what superview they belong to (page). The UI updates accordingly.
In order to determine whether the columnRange has overflowed from a container due to a page break OR whether the range of text hasn't overflowed its container and is actually using less space than available and should be sized down, I call both:
layoutManager.textContainer(forGlyphAt: lastGlyphOfColumn, effectiveRange: &actualGlyphRangeInContainer)`
// and
`layoutManager.lineFragmentRect(forGlyphAt: lastGlyphOfColumn, effectiveRange: nil)
Apple Documentation notes that both these calls force glyph generation and layout. As I'm in early development, I have not set non-contiguous layout. So these should be causing full layout, assuring accurate return values.
Or so I'd hoped.
This does work fine in many cases. I edit. Pagination works. But then I'll encounter UI-breaking inconsistent returns from these two calls. By inconsistent, I mean that the second call returns a line fragment rect that is in the container coordinates of A DIFFERENT container than the container returned by the first call. To be specific, the line fragment rect seems to be in the coordinates of the container that comes next in layoutManager.textContainers.
Example Code:
if !layoutManager.textContainers.indices.contains(i) {
containerToUse = createTextContainer(with: availableSize)
layoutManager.addTextContainer(containerToUse)
} else {
// We have a container already but it may be
// the wrong size.
containerToUse = layoutManager.textContainers[i]
if containerToUse.size.width != availableSize.width {
// Mandatory that we resize if we don't have
// a matching width. Height resizing is not
// mandatory and requires a layout check below.
containerToUse.size = availableSize
}
}
let glyphRange = layoutManager.glyphRange(forCharacterRange: remainingColumnRange, actualCharacterRange: nil)
let lastGlyphOfColumn = NSMaxRange(glyphRange) - 1
var containerForLastGlyphOfColumn = layoutManager.textContainer(forGlyphAt: lastGlyphOfColumn, effectiveRange: &actualGlyphRangeInContainer)
if containerForLastGlyphOfColumn != containerToUse
&& containerToUse.size.height < availableSize.height {
// If we are here, we overflowed the container,
// BUT the container we overflowed didn't use
// the maximum remaining page space (this
// means it was a pre-existing container that
// needs to be sized up and checked once more).
// NOTE RE: THE BUG:
// at this point, prints show...
// containerToUse.size.height
// =628
// availableSize.height
// =648
containerToUse.size = availableSize
containerForLastGlyphOfColumn = layoutManager.textContainer(forGlyphAt: lastGlyphOfColumn, effectiveRange: &actualGlyphRangeInContainer)
}
// We now check again, knowing that the container we
// are testing flow into is the max size it can be.
if containerForLastGlyphOfColumn != containerToUse {
// If we are here, we have overflowed the
// container, so containerToUse size SHOULD be
// final/accurate, since it is fully used.
actualCharRangeInContainer = layoutManager.characterRange(forGlyphRange: actualGlyphRangeInContainer, actualGlyphRange: nil)
// Start of overflow range is the first character
// in the container that was overflowed into.
let overflowLoc = actualCharRangeInContainer.location
remainingColumnRange = NSRange(location: overflowLoc, length: remainingColumnRange.length - overflowLoc)
// Update page count as we have broken to a new page
currentPage += 1
} else {
// If we are here, we have NOT overflowed
// from the container. BUT...
// THE BUG:
// ***** HERE IS THE BUG! *****
lineFragmentRectForLastChar = layoutManager.lineFragmentRect(forGlyphAt: lastGlyphOfColumn, effectiveRange: nil)
let usedHeight = lineFragmentRectForLastChar.maxY
// BUG: ^The lines of code above return a
// fragment rect that is in the coordinates
// of the WRONG text container. Prints show:
// usedHeight
// =14
// usedHeight shouldn't be just 14 if this is
// the SAME container that, when it was 628
// high, resulted in text overflowing.
// Therefore, the line fragment here seems
// to be in the coordinates of the ENSUING
// container that we overflowed INTO, but
// that shouldn't be possible, since we're in
// a closure for which we know:
//
// containerForLastGlyphOfColumn == containerToUse
//
// If the last glyph container is the container
// we just had to size UP, why does the final
// glyph line fragment rect have a maxY of 14!?
// Including ensuing code below only for context.
if usedHeight < containerToUse.size.height {
// Adjust container size down to usedRect
containerToUse.size = CGSize(width: containerToUse.size.width, height: usedHeight)
} else if usedHeight == availableSize.height {
// We didn't force break to a new page BUT
// we've used exactly the height of our page
// to layout this column range, so need to
// break to a new page for any ensuing text
// columns.
currentPage += 1
} else if usedHeight > containerToUse.size.height {
// We should have caught this earlier. Text
// has overflowed, but this should've been
// caught when we checked
// containerForLastGlyphOfColumn !=
// containerToUse.
//
// Note: this error has never thrown.
throw PaginationError.unknownError("Oops.")
}
}
Per my comments in the code block above, I don't understand why the very same text container that just overflowed and so had to be sized up from 628 to 648 in order to try to fit a glyph would now report that same glyph as both being IN that same container and having a line fragment rect with a maxY of just 14. A glyph couldn't fit in a container when it was 628 high, but if I size it up to 648, it only needs 14?
There's something very weird going on here. Working with NSLayoutManager is a bit of a nightmare given the unclear documentation.
Any help or insight here would be massively, massively appreciated.
I'm trying to update my app to use TextKit 2. The one thing that I'm still not sure about is how I can get the selection frame. My app uses it to auto-scroll the text to keep the cursor at the same height when the text wraps onto a new line or a newline is manually inserted. Currently I'm using NSLayoutManager.layoutManager!.boundingRect(forGlyphRange:in:).
The code below almost works. When editing the text or changing the selection, the current selection frame is printed out. My expectation is that the selection frame after a text or selection change should be equal to the selection frame before the next text change. I've noticed that this is not always true when the text has a NSParagraphStyle with spacing > 0. As long as I type at the end of the text, everything's fine, but if I insert some lines, then move the selection somewhere into the middle of the text and insert another newline, the frame printed after manually moving the selection is different than the frame before the newline is inserted. It seems that the offset between the two frames is exactly the same as the paragraph style's spacing. Instead when moving the selection with the arrow key the printed frames are correct.
I've filed FB17104954.
class ViewController: NSViewController, NSTextViewDelegate {
private var textView: NSTextView!
override func loadView() {
let scrollView = NSScrollView(frame: CGRect(x: 0, y: 0, width: 400, height: 400))
textView = NSTextView(frame: scrollView.frame)
textView.autoresizingMask = [.width, .height]
textView.delegate = self
let paragraphStyle = NSMutableParagraphStyle()
paragraphStyle.lineSpacing = 40
textView.typingAttributes = [.foregroundColor: NSColor.labelColor, .paragraphStyle: paragraphStyle]
scrollView.documentView = textView
scrollView.hasVerticalScroller = true
view = scrollView
}
func textView(_ textView: NSTextView, shouldChangeTextIn affectedCharRange: NSRange, replacementString: String?) -> Bool {
print("before", selectionFrame.maxY, selectionFrame)
return true
}
func textDidChange(_ notification: Notification) {
print("after ", selectionFrame.maxY, selectionFrame)
}
func textViewDidChangeSelection(_ notification: Notification) {
print("select", selectionFrame.maxY, selectionFrame)
}
var selectionFrame: CGRect {
guard let selection = textView.textLayoutManager!.textSelections.first?.textRanges.first else {
return .null
}
var frame = CGRect.null
textView.textLayoutManager!.ensureLayout(for: selection)
textView.textLayoutManager!.enumerateTextSegments(in: selection, type: .selection, options: [.rangeNotRequired]) { _, rect, _, _ in
frame = rect
return false
}
return frame
}
}
It seems that NSTextView has an issue with deleting text and setting any attribute at the same time, when it also has a textContainerInset.
With the code below, after 1 second, the empty line in the text view is automatically deleted and the first line is colored red. The top part of the last line remains visible at its old position. Selecting the whole text and then deselecting it again makes the issue disappear.
Is there a workaround?
I've created FB16897003.
class ViewController: NSViewController {
@IBOutlet var textView: NSTextView!
override func viewDidAppear() {
textView.textContainerInset = CGSize(width: 0, height: 8)
let _ = textView.layoutManager
textView.textStorage!.setAttributedString(NSAttributedString(string: "1\n\n2\n3\n4"))
textView.textStorage!.addAttribute(.foregroundColor, value: NSColor.labelColor, range: NSRange(location: 0, length: textView.textStorage!.length))
DispatchQueue.main.asyncAfter(deadline: .now() + 1) { [self] in
textView.selectedRange = NSRange(location: 3, length: 0)
textView.deleteBackward(nil)
textView.textStorage!.beginEditing()
textView.textStorage!.addAttribute(.foregroundColor, value: NSColor.red, range: NSRange(location: 0, length: 2))
textView.textStorage!.endEditing()
}
}
}
demo code :
- (void)drawRect:(CGRect)rect {
CGContextRef context = UIGraphicsGetCurrentContext();
// Flip the coordinate system
CGContextSetTextMatrix(context, CGAffineTransformIdentity);
CGContextTranslateCTM(context, 0, self.bounds.size.height);
CGContextScaleCTM(context, 1.0, -1.0);
NSDictionary *attrs = @{NSFontAttributeName: [UIFont systemFontOfSize:20],
NSForegroundColorAttributeName: [UIColor blueColor],
NSUnderlineStyleAttributeName: @(NSUnderlineStyleThick),
};
// Make an attributed string
NSAttributedString *attributedString = [[NSAttributedString alloc] initWithString:@"Hello CoreText!" attributes:attrs];
CFAttributedStringRef attributedStringRef = (__bridge CFAttributedStringRef)attributedString;
// Simple CoreText with CTFrameDraw
CTFramesetterRef framesetter = CTFramesetterCreateWithAttributedString(attributedStringRef);
CGPathRef path = CGPathCreateWithRect(self.bounds,NULL);
CTFrameRef frame = CTFramesetterCreateFrame(framesetter,CFRangeMake(0, 0),path,NULL);
//CTFrameDraw(frame, context);
// You can comment the line 'CTFrameDraw' and use the following lines
// draw with CTLineDraw
CFArrayRef lines = CTFrameGetLines(frame);
CGPoint lineOrigins[CFArrayGetCount(lines)];
CTFrameGetLineOrigins(frame, CFRangeMake(0, 0), lineOrigins);
for (int i = 0; i < CFArrayGetCount(lines); i++) {
CTLineRef line = CFArrayGetValueAtIndex(lines, i);
CGContextSetTextPosition(context, lineOrigins[i].x, lineOrigins[i].y);
// CTLineDraw(line, context);
// You can comment the line 'CTLineDraw' and use the following lines
// draw with CTRunDraw
// use CTRunDraw will lost some attributes like NSUnderlineStyleAttributeName,
// so you need draw it by yourself
CFArrayRef runs = CTLineGetGlyphRuns(line);
for (int j = 0; j < CFArrayGetCount(runs); j++) {
CTRunRef run = CFArrayGetValueAtIndex(runs, j);
CTRunDraw(run, context, CFRangeMake(0, 0));
}
}
}
this code will use CTRunDraw to draw the content , and the underline will draw and show normally in iOS17 & Xcode 15 , But when you build it with XCode16 & iOS18 beta . the underline will be missing .