THIS ... IS ... FORTRESS!

Question

Fortress was a language being developed by the Sun Programming Language Research Group (R.I.P. Fortress) that had a unique property to it, it was possible to render ("Fortify") programs in different font-styles (i.e. blackboard bold, bold, italics, roman, etc. ). The goal is to represent a one-char Fortress variable in HTML markup.

Here's how the fortification of one-char variables worked (simplified/modified from documentation for code-golfing purposes):

• If the variable is a repeated capital (i.e. ZZ), it becomes formatted in blackboard bold (𝔸𝔹ℂ𝔻𝔼𝔽𝔾ℍ𝕀𝕁𝕂𝕃𝕄ℕ𝕆ℙℚℝ𝕊𝕋𝕌𝕍𝕎𝕏𝕐ℤ)
• If the variable is preceded by an underscore, the variable is rendered in roman font (left alone)
• If the variable is followed by an underscore, the variable is rendered in bold font (<b>v</b>)
• If the variable is neither preceded nor followed by an underscore, the variable is rendered in italic font (<i>v</i>)
• The codepoints of the blackboard bolds are: 𝔸:1D538, 𝔹:1D539, ℂ:2102, 𝔻:1D53B, 𝔼:1D53C, 𝔽:1D53D, 𝔾:1D53E, ℍ:210D, 𝕀:1D540, 𝕁:1D541, 𝕂:1D542, 𝕃:1D543, 𝕄:1D544, ℕ:2115, 𝕆:1D546, ℙ:2119, ℚ:211A, ℝ:211D, 𝕊:1D54A, 𝕋:1D54B, 𝕌:1D54C, 𝕍:1D54D, 𝕎:1D54E, 𝕏:1D54F, 𝕐:1D550, ℤ:2124. These count as one byte each in your program (if your language of choice can handle these characters at all)

Input will be either a repeated ASCII capital, or a single ASCII letter with either no underscore, a leading underscore, or a trailing underscore (AKA _a_ will not be an input). This is code-golf so lowest byte count wins!

Test cases:

a => <i>a</i>
BB => 𝔹
c_ => <b>c</b>
_d => d
E => <i>E</i>
G_ => <b>G</b>
_H => H
ZZ => ℤ

Links: Specification, Direct download of version 0.1 alpha.

Reference implementation (This would be in Fortress, but Fortress doesn't like most of the doublestruck characters, so this implementation is in D):

dstring fortify(string arg) {
    import std.string, std.conv;

    alias D = to!dstring; //Convert to a string that accepts the unicode needed
    dstring BB = "𝔸𝔹ℂ𝔻𝔼𝔽𝔾ℍ𝕀𝕁𝕂𝕃𝕄ℕ𝕆ℙℚℝ𝕊𝕋𝕌𝕍𝕎𝕏𝕐ℤ"d; //blackboard bold capitals
    string UC = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"; //normal ASCII capitals

    if(arg.length == 1)
        return D("<i>" ~ arg ~ "</i>");
    if(arg[0] == a[1])
        return D(BB[UC.indexOf(arg[0])]);
    if(arg[0] == '_')
        return D(arg[1]);
    return D("<b>" ~ arg[0] ~ "</b>");
}

Answer 1

Python 3.6, 159 131 128 bytes

1 byte saved thanks to @Zacharý

3 bytes saved thanks to @VаlueInk

28 bytes saved thanks to @Rod

lambda s:len(s)<2and f"<i>{s}</i>"or s[0]==s[1]and"𝔸𝔹ℂ𝔻𝔼𝔽𝔾ℍ𝕀𝕁𝕂𝕃𝕄ℕ𝕆ℙℚℝ𝕊𝕋𝕌𝕍𝕎𝕏𝕐ℤ"[ord(s[0])-65]or[f"<b>{s[0]}</b>",s[1]][s[0]=='_']

Try it online!

Answer 2

Ruby, 104 106 105+1 = 105 107 106 "bytes"

Probably works even better in Retina. Uses -p flag.

-1 byte from Zacharý.

gsub /^.$/,'<i>\0</i>'
gsub(/(.)\1/){$1.tr"A-Z","𝔸𝔹ℂ𝔻-𝔾ℍ𝕀-𝕄ℕ𝕆ℙℚℝ𝕊-𝕐ℤ"}
gsub /(.)_/,'<b>\1</b>'
gsub ?_,''

Try it online!

Answer 3

Jelly, 73 bytes

_{Turns out that not being able to use the BBB letters in the code is quite expensive.}

5ŀ”i
“Ðñṡ’Dẋ@€“¡ḞḄ’ż“¿¿Æ€¢¬µ‘+⁽ø³F⁸ṪO¤+ị¥Ọ
Ṫ
Ḣ5ŀ”b
;j“<>/”ṃ@“¢ʠf’
i”_+LµĿ

A full program taking one argument and printing the result.

Try it online! or see the test suite.

How?

The main entry point is the last line of code ("Main link").

5ŀ”i - Link 1: list of characters, s (length 1 & no underscore)
  ”i - literal character 'i'
5ŀ   - call link 5 as a dyad with s on the left and 'i' on the right

“Ðñṡ’Dẋ@€“¡ḞḄ’ż“¿¿Æ€¢¬µ‘+⁽ø³F⁸ṪO¤+ị¥Ọ - Link 2: list of characters, s (length 2 & no underscore)
“Ðñṡ’                                 - base 250 literal              1007245
     D                                - to decimal list               [1,0,0,7,2,4,5]
         “¡ḞḄ’                        - base 250 literal              111673
      ẋ@€                             - repeat with reversed @rguments for €ach -> [[111673],[],[],[111673,111673,111673,111673,111673,111673,111673],[111673,111673],[111673,111673,111673,111673],[111673,111673,111673,111673,111673]]
               “¿¿Æ€¢¬µ‘              - code page index list          [11,11,13,12,1,7,9]
              ż                       - zip together                  [[111673,11],[11],[13],[[111673,111673,111673,111673,111673,111673,111673],12],[[111673,111673],1],[[111673,111673,111673,111673],7],[[111673,111673,111673,111673,111673],9]]
                         ⁽ø³          - base 250 literal              8382
                        +             - addition (vectorises)         [[120055,8393],[8393],[8395],[[120055,120055,120055,120055,120055,120055,120055],8394],[[120055,120055],8383],[[120055,120055,120055,120055],8389],[[120055,120055,120055,120055,120055],8391]]
                            F         - flatten                       [120055,8393,8393,8395,120055,120055,120055,120055,120055,120055,120055,8394,120055,120055,8383,120055,120055,120055,120055,8389,120055,120055,120055,120055,120055,8391]
                                ¤     - nilad followed by link(s) as a nilad:                                                                                    ^
                             ⁸        -   chain's left argument, s  e.g.    "CC"                                                                                 |
                              Ṫ       -   tail (last character)             'C'                                                                                  |
                               O      -   cast to ordinal                   67                                                                                   |
                                   ¥  - last two links as a dyad:                                                                                                |
                                  ị   -   index into (1-indexed & modular)  8383 (this is at index 67%26=15 -----------------------------------------------------+ )
                                 +    -   add the ordinal                   8450
                                    Ọ - convert from ordinal to character   'ℂ'

Ṫ - Link 3: list of characters, s (length 2 & underscore at index 1)
Ṫ - tail (get the first character

Ḣ5ŀ”b - Link 4: list of characters, s (length 2 & underscore at index 2)
Ḣ     - head s (the non-_ character)
   ”b - literal character 'b'
 5ŀ   - call link 5 as a dyad with the non-_ character on the left and 'b' on the right

;j“<>/”ṃ@“¢ʠf’ - Link 5, wrap in a tag: element, tagName      e.g. 'a', 'i'
;              - concatenate the element with the tagName          "ai"
  “<>/”        - literal list of characters                        "<>/"
 j             - join                                              "a<>/i"
         “¢ʠf’ - base 250 literal                                  166603
       ṃ@      - base decompression with reversed @rguments
               -   "a<>/i" is 5 long, so 166603 is converted to
               -   base 5 [2,0,3,1,2,4,0,3] with digits "a<>/i"    "<i>a</i>"

i”_+LµĿ - Main link: list of characters, s (as specified only):
 ”_     - literal '_'
i       - index of '_' in s (1-indexed; 0 if not found)
    L   - length of s
   +    - addition
     µĿ - call link with that number as a monad with argument s
        - implicit print

Answer 4

JavaScript, 97 chars

([a,b])=>a==b?[...'𝔸𝔹ℂ𝔻𝔼𝔽𝔾ℍ𝕀𝕁𝕂𝕃𝕄ℕ𝕆ℙℚℝ𝕊𝕋𝕌𝕍𝕎𝕏𝕐ℤ'][a.charCodeAt()-65]:b?b=='_'?a.bold():b:a.italics()

Why a language have methods like String.prototype.italics and String.prototype.bold?

Thanks to Neil, save 9 bytes, use [...s] instead of s.match(/./u).