Lightly | A modern style for qt applications | Theme library

This is absolutely a Safari bug on mobile and desktop and it's been driving me nuts. Your image doesn't replicate the issue on desktop safari, but those with higher resolution and more obvious text do.

What's happening is that Safari now OCRs images to allow users to select and copy text from them. It does this by adding shadow DOM elements containing the text that float above the image, all contained within one DIV that entirely covers the image. This doesn't happen instantly, but seems to take a little time and perhaps depend on user interaction with the image.

You can watch this happen by opening the MDN web docs page for the map element here and looking at the Example (not the demo) with Inspect Element. The img tag with the map starts out simple, but once it shows that it has children you can expand it and see the shadow DOM elements. Once those are added the map no longer functions, presumably because the overlay elements pass through clicks to the image but not the map.

This is a hard problem to solve in general, and even a fairly hard problem to solve in specific cases.

I think the difference arose between version 1.1.27.1 and 1.1.28, probably from this NEWS item:

construction of interacting factors (e.g. when f1:f2 or f1/f2 occur in random effects terms) is now more efficient for partially crossed designs (doesn't try to create all combinations of f1 and f2) (GH #635 and #636)

My guess is that this changes the ordering of the components in the Z matrix, which in turn means that results of various linear algebra operations are not identical (e.g. floating point arithmetic is not associative, so while binary addition is commutative (a + b == b + a), left-to-right evaluation of a sum may not be the same as right-to-left evaluation ((a+b) + c != a + (b+c)) ...)

My attempt at reproducing the problem uses the same version of R ("under development 2022-02-25 r81818") and compares only lme4 package versions 1.18.1 with 1.1.28.9000 (development); any upstream packages such as Rcpp, RcppEigen, Matrix use the same versions. (I had to backport a few changes from the development version of lme4 to 1.1.18.1 to get it to install under the most recent version of R, but I don't think any of those modifications would affect numerical results.)

I did the comparison by installing different versions of the lme4 package before running the code in a fresh R session. My results differed between versions 1.1.18.1 and 1.1.28 less than yours did (both fits were singular, and the relative differences in the theta estimates were of the order of 2e-7 — still greater than your desired 1e-8 tolerance but much smaller than 1e-4 ...)

The results from 1.1.18.1 and 1.1.27.1 were identical.

• Q1: Why are your results more different between versions than mine?
  • in general/anecdotally, numerical results on Windows are slightly more unstable/differ more from other platforms
  • there are more differences between your two test platforms than among mine: R version, upstream packages (Matrix/Rcpp/RcppEigen/minqa), possibly the compiler versions and settings used to build everything [all of which could make a difference]
• Q2: how should one deal with this kind of problem?
  • as a minor frame challenge, why (other than not understanding what's going on, which is a perfectly legitimate reason to be concerned) does this worry you? The differences in the results are way smaller than the magnitude of statistical uncertainty, and differences this large are also likely to occur across different platforms (OS/compiler version/etc.) even for otherwise identical environments (versions of R, lme4, and other packages).
  • you could revert to version 1.1.27.1 for now ...
  • I do take the differences between 1.1.27.1 as a bug, of sorts — at the very least it's an undocumented change in the package. If it were sufficiently high-priority I could investigate the code changes described above and see if there is a way to fix the problems they addressed without breaking backward compatibility (in theory this should be possible, but it could be annoyingly difficult ...)

With the snippet below, with the -- outside:

Short answer, yes, it should be closed. And the reason can be found at this answer:

The JVM will release all active resources upon termination; however, this does not ensure that the other end will free the resource too, so explicitly closing resources is in every programmer's best interest.

So in my case, it is true that the EntityManager and factory are closed at application shutdown, but this does not ensure that they are properly dealt with on the other end. I didn't mention it in my question, but in fact the same thing holds true for the Mongo Client as well (see this answer):

If you ever re-deploy your web application without first restarting your application server, you must ensure that the MongoClient is closed when your web application is shutdown.

About the implementation I made an interface that I called DBInitializer. I instantiated an object of type MongoInitializer or MySQLInitializer (both implementing DBInitializer) inside the main method. See code for more clarity.

DBInitializer:

you can half the processing time by removing duplicate comparisons. I.e. once you compared "title1" and "title2", you don't need to compare "title2" and "title1" - you already know the answer. So, your inner loop should not start from the beginning of the array

As far as I can see this.getView().getConfig('fee_earner_user_id') is the only way when using the Architect.

The buffer should only be processed when the end event fires, otherwise you may be processing an incomplete buffer.

You could use json module to parse content of script tags, which is accessible by .text field

Here is the example of parsing all script jsons and printing name:

Lighting calculations should not be performed in clip space (including the projection matrix). Leave the projection away from all variables, including light positions etc., and you should be good.

Why is that? Well, lighting is a physical phenomenon that essentially depends on angles and distances. Therefore, to calculate it, you should choose a space that preserves these things. World space or camera space are two examples of angle and distance-preserving spaces (compared to the physical space). You may of course define them differently, but in most cases they are. Clip space preserves neither of the two, hence the angles and distances you calculate in this space are not the physical ones you need to determine physical lighting.