The current release doesn't support "multi-flags" like -abc. It would be nice if it did

I've just created a new branch for this issue. Anyone willing to help, please read this

The cmd

xorsum -a lo -A UP

Should output lowercase hex digest for the file "lo" and uppercase hex for "UP". Instead of the current behavior which uses whatever last capitalization is specified.

This gives more freedom to users, and makes the program "more correct"

Like b3sum does, this should allow users to set the digest size to any number of bytes in the interval [0, 256]

Very fundamental functionality

• Avoid crash if file doesn't exist, and print error message to stderr
• Avoid crash if path is a directory
• Avoid crash if there's no permission to read a file
• Handle possible std{in, out, err} stream errors
• Avoid panic if --length <n> cannot be parsed to usize (handled by clap)
• Return meaningful exit codes

I think there should be a single --code parameter that's internally an enum that only accepts hex-UP, hex-lo, hex-UP-alt, hex-lo-alt, raw, and possibly b64 (in the future). Maybe some "CSV" modes that add whitespace or hyphens between groups of nibbles.

This is more future-proof, and avoids invalid or useless OPTION combinations. It may also simplify the experience for users.

This would definitely require a new major semver bump

This is essential (not fundamental) functionality for checksum programs. It should be capable of verifying and comparing hash values from a file(s)

This algorithm is infinitely parallelizable, we should take advantage of that

That vector is always initialized to all 0s before being passed to that fn, so it makes sense to create it there and return it after mutating, instead of passing a mutable reference and manually resetting with fill

There are 3 instances of the XOR cipher in the code. This redundancy is unacceptable

Activating this mode would change the behavior to be inconsistent with core-utils, but with the bonus of having #13 and access to #15.

It should be activated by using the --quirky flag, no shorthand allowed (to enforce being explicit)

Currently, the default Initialization Vector is 0, but users should be capable of changing it

More info

Do the same thing as b3sum to ensure checkfiles can be shared between DOS and Unix users. This is necessary before implementing #6

The program doesn't lock stdout when in raw mode, but it does lock in hex mode. For the sake of consistency, it should be the same regardless of the mode.

This isn't important now, but it will be important when implementing #4

It should be the same as md5sum

It should return an empty hex string, not throw an error

I did some benchmarks, and this is CONSIDERABLY SLOWER than MD5 and Blake2. This is unacceptable. How come such a simple algorithm is slower than a single-threaded (I guess, didn't check) MD5?

• The first one to add would be

xorsum --hell
# out: "I can't go to hell. I'm all out of vacation days."

As an Undertale reference. Because typing --hell is a common typo when requesting help from a CLI program.

• Another related one would be:

xorsum --quirky -l 6 hell
# "666666666666 hell"

Regardless if the file exists or not, and regardless of its contents. The length of sixes corresponds to the specified value of -l. To get the ACTUAL hash, the user must be explicit about it:

xorsum --quirky -l 6 ./hell

• An old friend:

xorsum --hello
# "world!"

• Rickroll lol:

xorsum --rick
# "We're no strangers to love..."

• A Spamton reference:

xorsum --heaven
# "[Heaven], are you WATCHING?"

• My Hero Academia:

xorsum --quirky --stress
# "Doofenshmirtz Glowup moment"

Because of the meme about re-destro.

• Umm... what?

xorsum --quirky --quirky
# "Very quirky indeed..."

• This one would cause an infinite loop:

xorsum --quirky xorsum