fix: broken, infinitely looping minPresenceAhead calculation

[carlobeltrame]

Fixes #2303, where sometimes react-pdf could be caught in an infinite loop during rendering!
Fixes #2238

I debugged the existing shouldBreak logic which was introduced for react-pdf v2, and found some major flaws in the logic. At first I tried to just fix the logic, but I noticed it could be done more efficiently and cleanly, so I went ahead and rewrote part of the logic.

Problem 1: When calculating the "presence" of the sibling nodes following the child in question, vertical margin was completely ignored so far. I agree to ignore the bottom margin of the bottom-most node, but all other margins must be taken into account.

In this picture, shouldBreak(child, futureElements, ...) returned false so far because the height alone of futureElements[0] was not enough to exceed the page height. The marginTop o futureElements[0] was completely ignored in the old logic. The new logic correctly returns true in this case, because it searches for the last box end in all futureElements.

Problem 2: The previous logic would ignore the minPresenceAhead value, if the heights of the nodes following the child didn't add up to at least minPresenceAhead.

In this picture, the two futureElements have a height total of 80, or 90 if one includes the vertical margins except the last bottom margin, which is less than the 100 minPresenceAhead. The old logic would completely ignore the minPresenceAhead, when the futureElements aren't at least this high. The new logic correctly outputs true for shouldBreak(child, futureElements, ...) in this case.

Problem 3: The previous logic would in certain scenarios output true for shouldBreak(child, futureElements, ...) even when the child is already at the top of the page. Breaking the child over to the next page wouldn't improve the situation at all, because the breaking conditions would then be met again on the next page, and on the next, and so on, leading to an infinite loop. The new logic prevents this by detecting whether the child is already at the top of the page.

Problem 4: The previous logic would incorrectly take fixed futureElements into account. The new logic correctly ignores fixed elements, because they do not have any influence on the page flow.

Notes

• I completely removed the getNodesHeight function and file, because it was only used in shouldBreak, had no tests, was implemented in an overly specific way for just this one use case, and was not necessary anymore after my rewrite.
• I also removed the option to specify a presenceAhead function as a prop on any element. This was undocumented, I didn't find any internal uses and it did not seem very helpful to me in the current form (at least, the function would require element as an argument to be useful, like defaultPresenceAhead receives it).
• I added tests for 100% branch coverage (I hope), as well as regression tests for the known REPL case from the original issue minPresenceAhead crashed application #2303 and for a minimal reproduction example which I created while debugging this code.

[changeset-bot]

🦋 Changeset detected

Latest commit: ff69234

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 3 packages

Name	Type
@react-pdf/layout	Minor
@react-pdf/renderer	Patch
@react-pdf/examples	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[thedivac]

Great work! That fixed a bunch of issues I was having with my layout!

[carlobeltrame]

Great work! That fixed a bunch of issues I was having with my layout!

Great to hear someone other than me was finally able to test this!
@diegomura would you care to review? These problems have reportedly affected a number of users now.

[diegomura]

tACK

Amazing. Thanks so much!