More low-hanging transformations

[jvoisin]

Maths

• [x] Number(s) → ~+[s]~ +s
• [ ] n=n-1 → n--
• [ ] n=n+1 → n++
• [ ] all the n = n ° m → n°=m family, with ° being +, -, /, * and %
• [x] ~~ParseFloat(n) → +n~~
• [ ] x!=y → x^y if x and y are integers
• [ ] -10 → ~9, -100 → ~99, … and so on :D
• [x] N00…0 with N an integer → Ne00…0
• [ ] a>=b → a>b-1 if -10 < b < 10
• [ ] a<=b → a<b-1 if -10 < b < 10
• [x] Math.abs(a) → a>0?a:-a if a is one char.
• [ ] Math.sqrt(a) → a**.5

Arrays

• [ ] a.concat() → [...a]
• [ ] x[x.length-N] → x.at(N)

Random

• [ ] Math.random()<0.5 → new Date&1, as Date returns a number of ms, albeit it might be problematic if called in a super-tight loop.
• [ ] Math.floor(Math.random()*N)) → new Date%N

Misc

• [x] while(1) → for(;;)

[tdewolff]

Thanks Voisin. What's the difference between +[s] and +s?

[jvoisin]

Oops, it's a typo, +s should be enough for Number

[tdewolff]

The Random ones are not feasible. Math.random is not the same as new Date, the former is probably a much more secure random source, while the latter is a poor man's random source. I believe we already do the while(1) => for(;;) one.

The parseInt and parseFloat conversions are actually not such a good idea. Try inputting something invalid and the output is different between the two.

[jvoisin]

> parseFloat("aaa")
NaN
> +"aaa"
NaN 

looks identical do me.

[tdewolff]

Try "5a" for example, or try parseInt("5.5") != +"5.5"

On Tue, Jan 21, 2025, 12:02 Julien Voisin @.***> wrote:

parseFloat("aaa") NaN +"aaa" NaN

looks identical do me.

— Reply to this email directly, view it on GitHub https://github.com/tdewolff/minify/issues/794#issuecomment-2604405397, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABKOGHVJWUTU2PXMRO76XOT2LYSK3AVCNFSM6AAAAABVRBRLX6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMMBUGQYDKMZZG4 . You are receiving this because you commented.Message ID: @.***>

[jvoisin]

> +"5a"
NaN
> parseInt("5a")
5
> parseFloat("5a")
5 

Damn :<

[torstenvl]

I'm not sure n=n-1 → n-- is good either. What if I have x = y = n = n-1?

It's possible that --n works—I think it should, but I haven't tested it.

[tdewolff]

Yes, many of the conversions are incorrect, but converting (n = n + 1) => (++n) should be fine (not n++ as you correctly mention).

The ones that are valid and still missing are:

• [x] n=n+1 → ++n and n=n-1 → --n
• [x] all the n = n + m → n+=m for all operators (+, -, /, *, %, **, <<, >>, >>>, &, ^, |, &&, ||, ??)
• [ ] -10 → ~9, -100 → ~99, … and so on :D
• [x] Math.sqrt(a) → a**.5
• [ ] x[x.length-N] → x.at(N) (not sure if this is valid for all types that permit the index operator?)

[tdewolff]

I've implemented the n=n+1 => ++n, the n=n+a => n+=a, and Math.sqrt(a) => a**.5 transformations!