To interpolate or concatentate, that is the question

[JakeQZ]

Whether 'tis nobler in the mind to suffer
The slings and arrows of outrageously undreadable code,
Or to take arms against a sea of super-optimized code
And by opposing use a few more CPU cycles and CO2.

Following on from #1444 (comment)...

This probably isn't going anywhere, though I would like to note links to the various other discussions I found on the topic, which seem to concude:

• Interpolation is more optimal, increasingly so with the number of variables that need to be included in the string;
• Readability is in the eye of the beholder (it depends what you're used to seeing);
• Most IDEs are quite capable of syntax-highlighting the interpolated variables.

I did not find anything about how PHPStan handles interpolation, but I expect it does just fine at that (it's unexpectedly good at other analysis).

References:

• StackOverflow discussion about performance
  • Conclusion
    • no perceptible difference in performance
• General StackOverflow discussion
  • Conclusion
    • no perceptible difference in performance
    • matter of taste
• Answer to a StackOverflow qeustion, commenting on performance
  • Conclusion
    • interpolation is better for performance if there are many variables to insert
• StringConcatenationToInterpolation fixer PHP-CS-Fixer/PHP-CS-Fixer#2334
  • This is a request for the converse of what I was looking for, which I could not find
  • Conclusion
    • Developers overall seem to prefer interpolation
    • Interpolation performance is better by about 15% with two strings to insert
    • It's only harder to read the code if you are not used to it

My conclusion:

• We should continue to allow and use both on a case-by-case basis, and should not change existing code to use the other without good reason
• I would favour the use of interpolation for the sake of CO2 emissions as well as performance.

[oliverklee]

I'd be fine either way, and also fine with deciding on a case-by-case basis. (And generally, I think we should favor readability over micro-optimization of things that are not proven to be performance bottlenecks.)