Count the characters - bit by bit!

Question

The simple part: Given an input string containing only printable ASCII-characters (space - tilde), count the number of occurrences of each character and return the result on any convenient format. The result for a string a%hda7a should be something like: a:3, %:1, h:1, 7:1, d:1. Sorting is unnecessary, the delimiters and formats are optional but it must be easily understood which number corresponds to which character. You shall not include characters that are not in the input string (a:3, b:0, c:0, d:1, ... is not OK).

The real challenge:

Convert every character in your code to an 8-bit binary number (or 16-bit if you're using UTF-16 or similar), and enumerate every character starting at 0.

For every character (i is the enumerator), the i%7-bit¹ must be 1. The bits are numbered from the right. All other bits can be whatever you want.

Let's use the following code as an example:

[f]-xif)#f

Converting this to binary we get the array below. The first number (representing [ has a 1 in the 0'th position, so that one is OK. The second number (representing f has a 1 in the 1'st position, so that one is OK too. Continue like this, and you'll see that the code above is valid.

C  76543210  Bit number
-  --------  ----------
[  01011011  0   - OK
f  01100110  1   - OK
]  01011101  2   - OK
-  00101101  3   - OK
x  01111000  4   - OK
i  01101001  5   - OK
f  01100110  6   - OK
)  00101001  0   - OK
#  00100011  1   - OK
f  01100110  2   - OK

If we change the code into: ]f[-xif)#f we'll get the following start of the sequence:

C  76543210  Bit number
-  --------  ----------
]  01011101  0   <- OK
f  01100110  1   <- OK
[  01011011  2   <- Not OK
-  00101101  3   <- OK

As we see, the third character [ doesn't have a 1 in the 2nd position (zero-indexed), and this code is therefore not valid.

Test cases:

Input:
This is a string containing some symbols: ".#!".#&/#

Output:
   !  "  #  &  /  :  T  a  b  c  e  g  h  i  l  m  n  o  r  s  t  y  .
7  1  2  3  1  1  1  1  2  1  1  1  2  1  5  1  2  4  3  1  6  2  1  2

Any reasonable output format is OK (whatever is most convenient to you). You could for instance have: :7, !:1, ":2, #:3, &:1, /:1, T:1, a:2 ... or [ ,7][!,1][",2][#,3][&,1].... The output is on any standard way (return from function, printed to STDOUT etc.)

¹i modulus 7.

---

This is code-golf, so the shortest code in bytes will win^ref.

Answer 1

Pyke, 1 6 bytes

1cn;1c

Try it here!

1c     - chunk(size=1, input)
  n;1  - noop. 
     c - count(^)

Half of this code is just no-ops...

00110001 - 1
01100011 - c
01101110 - n
00111011 - ;
00110001 - 1
01100011 - c

Answer 2

Pyth, 12 8 7 bytes

^{-1 byte thanks to @Loovjo}

m+d/Qd{
      { # remove all duplicated elements from the (implicit) input
m       # map each element (d) of the parameter (the set from previous operation)
   /Qd  # count the occurrences of d in Q
 +d     # concatenate with d

binary representation

01101101 m
00101011 +
01100100 d
00101111 /
01010001 Q
01100100 d
01111011 {

Try here

Answer 3

Befunge-93, 150 bytes

={<{p+}3/}*77\%*7{7:\+{}{1g}+3/*77\%*7{7:}=:_{}{}={}{}{v#{}{}`x1:~
}-}=*}{2*}97}:<$}={$_v#}!:-*84g+3/*77\%*7{7:}=:}:}+}1{}<_{@#
}{}{}={}{}{}={^.},\={<

Try it online!

I started by writing this as a regular Befunge program, which I golfed as much as possible. I then added padding to make sure the various characters in the program only appeared in permitted positions. This padding relied on the fact that unsupported commands are ignored in Befunge-93, so I just needed a sequence of unused characters whose bits aligned with the positions required (the sequence I used was ={}{}{}).

The complicated bit was that the various branches between lines needed to line up correctly (e.g. the v arrow in one line, would need to line up with the < arrow below it). This was further complicated by the fact that the bridge command (#) could not be separated from its adjacent branching arrow. Initially I tried generating the padding programatically, but in the end it was mostly a manual process.

Because of the size of the program I'm not going to list the full character analysis, but this is a sample from the beginning and the end:

= 00111101 0
{ 01111011 1
< 00111100 2
{ 01111011 3
p 01110000 4
+ 00101011 5
} 01111101 6
3 00110011 0
/ 00101111 1
...
{ 01111011 1
^ 01011110 2
. 00101110 3
} 01111101 4
, 00101100 5
\ 01011100 6
= 00111101 0
{ 01111011 1
< 00111100 2

The line breaks are treated as a newline characters, so will either be in position 1 or 3.

Answer 4

MATL, 17 bytes

u"G91x@=zD91x@uRD

Displays the count, then the corresponding character, all newline-separated. Biggest difficulty is the @ which is 0b01000000; I hope I can find a way to make do without it.

Try it online!

Explanation:

u"  % Implicit input. Take (u)nique characters and loop (") over them.
G   % Take the input a(G)ain
91x % Filler: push 91, delete immediately.
@   % Push current character of loop
=   % Check for equality with earlier G
z   % Count number of equal characters
D   % Display
91x % More filler!
@   % Get loop character again
uR  % Filler: two NOPs for the single-character @
D   % Display. Implicitly end loop.

MATL, 15 bytes (questionable output)

If just leaving two row vectors on the stack is allowed (function-like behaviour as per this Meta post), we can get down to

u"G91x@=zv]v!Gu

But here, the output is not quite as neatly ordered.

Answer 5

CJam, 14 bytes

q__|_ @sfe=]zp

Try it here.

The space before the @ and the s after it are filler characters inserted to make the ASCII codes fit the required pattern: the space does nothing, and the s just converts a string into a string. Other than that, this is a pretty simple and straightforward implementation of the challenge task:

q_               "read the input and make a copy of it";
  _|             "collapse repeated characters in the copy"; 
    _            "save a copy of the collapsed string";
      @          "pull the original input string to the top of the stack";
       s         "(does nothing here)";
        fe=      "for each character in the collapsed string, count the ...";
                 "... number of times it occurs in the original string";
           ]z    "pair up the counts with the saved copy of the collapsed string";
             p   "print the string representation of the result";

For the input foobar123, this code outputs [['f 1] ['o 2] ['b 1] ['a 1] ['r 1] ['1 2] ['2 2] ['3 1]]. If simply printing the counts on one line and the corresponding characters on another, as in:

[1 2 1 1 1 2 2 1]
fobar123

is considered an acceptable output format, then the ]z may be omitted to save two bytes, for a total of 12 bytes. Yes, the shortened code will still pass the bit pattern requirement.

Ps. I also wrote a simple source code checker for this challenge. Given a line of code as input, it will first echo that line and then print the same line with each character replaced by its (n % 7)-th ASCII bit. If the second line is all ones, the input is valid.

Answer 6

Jelly, 6 bytes in Jelly's codepage

ṢZṢṀŒr

Try it online!

This is a function that returns a list of (character, count) pairs. (Jelly represents such lists as text, e.g. if they're sent to standard output, by concatenating the elements, which is why you have to treat this as a function rather than a complete program. (Here's the same program with some code appended to call the function and then print the internal structure to standard output, proving that the output's in an unambiguous format.)

Binary representation and explanation:

  76543210 

Ṣ 10110111   Sort the characters of the input
Z 01011010   Transpose the list (it's 1D, so this effectively wraps it in a list)
Ṣ 10110111   Sort the list (a no-op, as it has only one element)
Ṁ 11001000   Take the largest (i.e. only) element
Π00010011   First byte of a two-byte command
r 01110010   Run-length encode

It can be seen that the second, third, and fourth characters cancel each other out and are only there to maintain the bit pattern we need. Œr is just too convenient, though, and padding the program so that we can use it probably gives us a shorter program than trying to solve the problem without the builtin would.