I need to implement a RegexMathEngine optimization that treats (?!p)q as (?=q)(?!p)q for expressions where it can know q will evaluate faster. But I've got to figure out how to show it in the debug trace.

And thanks :)

Found a big optimization for smallest factor of N sharing all prime factors with N.

That’s a weird way to say “product of prime factors of N”

Done reading through it, gratz on the optimization!

Thanks :) I would really like to replace it with another answer that requires that virtualized-remainder-math trick, and I think your perfect powers problem would do nicely as a question that isn't just specialized to regexes. The numbers need a name though :) something referring to the perfect power hierarchy they contain.

Balance of power numbers? :)

You mean the one where A, B are powers, (N-A) is a power of (N-B)?

Yep, that's the one.

-1 char and -1 group on “find the biggest prime factor”

(?=((xx+)\2*(?=\2$)|x*)*)

And another -2 with (?=((xx+)(?=\2+$)|x*)*)

Can drop the |x* for -3 if it doesn’t need to work for prime numbers

But special-casing primes costs me +5, so not worth it

@Grimy Very nice! I would put a + in place of the last *

Done that

That causes it to fail for 1, saving another char

Because then we can change x$| at start to |x at end

@Grimy That's not what I meant... and no, it doesn't? You need xx+ to do that, don't you?

Oh I meant the very last star

That doesn't make it fail for 1 either

Oh you mean the main loop?

Yeah

So you're at 131 now, right? And 120 with (?*)? (I haven't yet looked for where to do that)

131, 127 with (?*)

hah, 40 strikes again

"Also I just got 130 with (?*)" <- so this was after you'd already gone lower than 141 without (?*)?

Well yeah, we were discussing a 137 right before that

Oops, misremembered. Yeah, I had the 136 at that point when you said that

But 136-130 = 6, and 131-127=4

Non-(?*) had 2 \10, (?*) had none

So only the former benefited from saving a capture group

Ah

@Grimy How can that be? There was just one \10 at that point

As there is now just one \9

Oh, I counted the \9 in the comment

I dunno where the other char disappeared, then

Holy shit, just started analyzing the rest of your totient regex... ((x*)(?=\5\9*$)x) did you just optimize this form of division?!

This is huuuuge

why didn't you say something

this will drop bytes in many of my other regexes

Will it? I didn’t realize, that’s cool.

Aw damnit, github just broke my workaround for making things fill the width of the browser instead of being limited to a really narrow width.

.container { margin-left:auto; margin-right:auto; width:2882px !important; } - this doesn't work anymore

I just click raw x)

For gists?

gist.github.com/Davidebyzero/9221685

I designed this to require my workaround to be viewed in a sensible way

Oh not for gists no

This looks fine for me, except the “Robust solution” table requires h scrolling

For example "Robust solutions for Triples" is supposed to show each regex as a single unwrapped line

so you can at a glance see the differences in their length

But you're seeing them wrapped to up to 6 lines, right?

Right

Ah, just have to apply the same to .container-lg

Here’s my current 131: pastebin.com/raw/yqpnzrwj. I’m sure there’s a way to save on that \5 computation, but I’m heading off to bed for now.

I replicated the 131 already, but it's the 127 I haven't

Goodnight :)

(When I said "So you're at 131 now, right?" I was measuring the length of a working regex, not applying the -2 -3 stuff you said.)