Python2, 55 Bytes

a=" )dfi(+m,nprut.';";print map(('a="'+a).find,input())

Starts with a string that contains all the characters used in the code, and then search the indexes

Jelly, 10 9 bytes

“ṾiЀƓv”v

Try it online!

How it works

“ṾiЀƓv”v  Main link. No arguments.

“ṾiЀƓv”   Set the left argument and the return value to s := 'ṾiЀƓv'.
        v  Execute the string s as a monadic Jelly program with argument s.

 Ṿ         Uneval; yield a string representation of s, i.e., r := '“ṾiЀƓv”'.
     Ɠ     Read one line from STDIN and evaluate it like Python would.
  iЀ      Find the index of each character in the input in r.
      v    Eval the list of indices as a monadic Jelly program with argument s.
           Why?
             This is the shortest way to add the character 'v' to the string s,
             meaning that we can use r without having to append anything.
           What?
             The v atom vectorizes at depth 1 for its left argument, meaning that
             it acts on arrays of numbers and/or characters. When fed an array of
             integers, it first converts them to strings, then concatenates the
             strings and evaluates them as a Jelly program. For example, the array
             [1, 2, 3] gets cast to the string '123', then evaluates, yielding 123.
             Something slightly different happens if the array starts with a 0. For
             example, the array [0, 1, 2] gets cast to '012' just as before, but
             Jelly views '0' and '12' as two separate tokens; numeric literals
             cannot start with a 0. Since the Jelly program is monadic, the first
             token – '0' – sets the return value to 0. Since the second token –
             '12' – is also a niladic link, the previous return value is printed
             before changing the return value to 12. Then, the program finishes
             and the last return value is printed implicitly.

Lenguage, 56,623 bytes

Below is a hexdump of the first 256 bytes. The remaining bytes can be chosen arbitrarily.

0000000: 00 01 02 03 04 05 06 07 08 09 0a 0b 0c 0d 0e 0f  ................
0000010: 10 11 12 13 14 15 16 17 18 19 1a 1b 1c 1d 1e 1f  ................
0000020: 20 21 22 23 24 25 26 27 28 29 2a 2b 2c 2d 2e 2f   !"#$%&'()*+,-./
0000030: 30 31 32 33 34 35 36 37 38 39 3a 3b 3c 3d 3e 3f  0123456789:;<=>?
0000040: 40 41 42 43 44 45 46 47 48 49 4a 4b 4c 4d 4e 4f  @ABCDEFGHIJKLMNO
0000050: 50 51 52 53 54 55 56 57 58 59 5a 5b 5c 5d 5e 5f  PQRSTUVWXYZ[\]^_
0000060: 60 61 62 63 64 65 66 67 68 69 6a 6b 6c 6d 6e 6f  `abcdefghijklmno
0000070: 70 71 72 73 74 75 76 77 78 79 7a 7b 7c 7d 7e 7f  pqrstuvwxyz{|}~.
0000080: 80 81 82 83 84 85 86 87 88 89 8a 8b 8c 8d 8e 8f  ................
0000090: 90 91 92 93 94 95 96 97 98 99 9a 9b 9c 9d 9e 9f  ................
00000a0: a0 a1 a2 a3 a4 a5 a6 a7 a8 a9 aa ab ac ad ae af  ................
00000b0: b0 b1 b2 b3 b4 b5 b6 b7 b8 b9 ba bb bc bd be bf  ................
00000c0: c0 c1 c2 c3 c4 c5 c6 c7 c8 c9 ca cb cc cd ce cf  ................
00000d0: d0 d1 d2 d3 d4 d5 d6 d7 d8 d9 da db dc dd de df  ................
00000e0: e0 e1 e2 e3 e4 e5 e6 e7 e8 e9 ea eb ec ed ee ef  ................
00000f0: f0 f1 f2 f3 f4 f5 f6 f7 f8 f9 fa fb fc fd fe ff  ................

Output is in bytes, as customary for brainfuck et al.

How it works

This is a simple cat program, specifically ,[.,].

The source code contains all 256 byte values in order, so each byte's index in it matches its value.

CJam, 14 12 bytes

{sq\f#p_~}_~

Uses 0-based indexing and -1 for characters that don't appear in the source.

Try it online!

Javascript, 34 bytes

f=a=>a.map(v=>('f='+f).indexOf(v))

It takes input as array of strings, x is -1 (0-based indexing).

C, 153 152 143 bytes

char s[99],p[]="odeflnrti%()*+-0;<={}\\";c;f(char*i){sprintf(s,"char s[99],p[]=\"%s",p);for(c=0;c<strlen(i);)printf("%d ",strchr(s,i[c++])-s);}

Try it online!

Ruby, 41 88 86 71 69 67 61 bytes

a='p$<.chrsm{| #index};"';p$<.chars.map{|c|"a='#{a}".index c}

Thx Lynn for killing 6 bytes

><> (Fish) 70 bytes

 #.0+4*a5;!?l|!?f4*b+l1--naolc3*1+0.01?!|~ed+0.0+2e-{:;!?+1:i-1:r}+2:"

Probably the longest ><> 1 liner I've ever made.

It will print the output for each character found on a separate line (0 indexed).

A non found character will always print the length of the code + 1 (I could change this if deemed not okay in it's current state) so in this case 71 will always be the "Not found" characters.

I'll run up an explanation once I get the time.

Some test cases;

##K = 1\n1\n71

#"# = 1\n69\n1

Try it online

><> language

Lenguage, 1.22e7 bytes

Consists of 12263215 NUL bytes, (Hex 0x00).

Outputs a NUL for every character that doesn't appear in the source.

The ruse is that the input will never contain a NUL, so we always output the amount of NULs that there are characters in the input.

This translates to the following Brainfuck program

,[[-].,]

And with a breakdown...

,[[-].,]
,[    ,]    #Basic Input loop.
  [-]       #Zero out the cell.
     .      #Print it (A NUL).

This just shows the sheer power of Lenguage as a golfing language. Fear it.

Perl 6, 50 52 bytes

{(('R~.index$_) for}\\'R~'{((\'').index($_) for $_)}

Translation of G B's Ruby solution and Rod's Python solution.

A lambda that inputs a list of characters and outputs a list of zero-based indexes (Nil for nonexistent characters).

EDIT: Fixed an oversight - required adding 2 bytes :(

Stacked, noncompeting, 36 bytes

When I said this language was still in development, I meant it. Apparently, prompt used to consume the entire stack. This is why I can't have nice things. Try it here!

[tostr ':!' + prompt CS index out]:!

This is the standard quine framework. Basically, : duplicates the function [...] on the stack, which is then executed with !. Then, the inside of [...] executes with the function on the stack. It casts it to a string, appends :! (the program itself), then takes a string input with prompt. CS converts it to a character string. A character string is a bit different from a regular string in that it has operators vectorize over it. In this case, index vectorizes over the input, yielding each index of the input string in the program, finally being outputted.

For input Hello, World!, this gives:

(-1 27 -1 -1 2 -1 6 -1 2 5 -1 26 9)

I tried using the one without a quine (i.e. encoding the string of characters that appear in your source), but there is only one type of quotation mark in Stacked, namely, ', so it would be longer to do that type of solution.

Clojure, 43 56 bytes

Edit: Damn I forgot about 2! Increased from 43 to 56.

{\{ 0\\   1\  3\0   4\1 10\4 20\3 14 \} 42}

The hash-map consists only of characters 01234{\\}, and it encodes their locations. In Clojure hash-maps can be used functions, as shown in this complete example (f could be replaced by the hash-map definition):

; Keeping track of the zero-based index:
;      00000000001111111111222222222233333333334444444444
;      01234567890123456789012345678901234567890123456789
(def f {\{ 0\\   1\  3\0   4\1 10\3 14\4 20\2 34 \} 43})

(map f "0123456789{} \\abcdef") ; (4 10 34 14 20 nil nil nil nil nil 0 43 3 1 nil nil nil nil nil nil)
(apply str (keys f))            ; " 01234{\\}"

I guess this counts :)

Python, 90 88 bytes

a,b,d=" ()+.7:[]efilmnor","a,b,d=\"",lambda e:[[b.find(d),a.find(d)+7][d in a]for d in e]

Test case:

print(d("a,b(]q"))
#[0, 1, 2, 8, 15, -1]

Pyth, 11 bytes

xL_+N"N+_Lx

A program that takes input of a "quoted string", with any quotes in the string escaped with a preceding \, and prints a list of zero-indexed values with -1 for characters not in the source.

Try it online!

How it works

xL_+N"N+_Lx    Program. Input: Q
xL_+N"N+_Lx"Q  Implicit quote closure and implicit input
     "N+_Lx"   Yield the string "N+_Lx"
   +N          Prepend a quote
  _            Reverse
 L          Q  Map over Q:
x               Yield the index of the character in the string
               Implicitly print