nandi/frq · a16dfae

nandi/frqpublic Fork

⑂ a16dfae

Undo the escaping on a tag value before reading it

IRCv3 escapes tag values, because `;` separates tags and a space ends them:
a semicolon travels as `\:`, a space as `\s`. `tag-value` took values as they
came, which was fine for the two things it was written for — a msgid and a
timestamp contain neither character — and wrong for anything else.

A reaction tally is one of the anything else. `emoji:nick;emoji:nick` arrives
with every separator written `\:`, so a reader that skips this step sees one
tally where there were three, and counts to match: the first emoji swallowing
the rest of the line as its list of nicks.

Values are unescaped now, `\\`, `\r` and `\n` included.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

nandi committed 2026-08-30T12:26:16-07:00 Browse files

a16dfae parent: ef6123f

modified src/frq/irc.jolt +26 -3 Viewed

@@ -73,14 +73,37 @@
73	73	:command (str/upper-case (or (first parts) ""))
74	74	:params (cond-> (vec (rest parts)) trailing (conj trailing))}))
75	75
	76	+(defn unescape-tag
	77	+ "An IRCv3 tag value with its escapes undone.
	78	+
	79	+ `\\:` is a semicolon, `\\s` a space, and `\\\\`, `\\r` and `\\n` themselves —
	80	+ the escaping exists because `;` separates tags and a space ends them. It
	81	+ matters for any value that can contain either: a reaction tally is
	82	+ `emoji:nick;emoji:nick` on the wire and arrives with every one of those
	83	+ semicolons written `\\:`, so a reader that skips this step sees one tally
	84	+ where there were three, and counts to match."
	85	+ [v]
	86	+ (when v
	87	+ (loop [in (seq v) out []]
	88	+ (if-let [c (first in)]
	89	+ (if (and (= \\ c) (second in))
	90	+ (recur (drop 2 in)
	91	+ (conj out (case (second in)
	92	+ \: \;
	93	+ \s \space
	94	+ \r \return
	95	+ \n \newline
	96	+ (second in))))
	97	+ (recur (rest in) (conj out c)))
	98	+ (apply str out)))))
	99	+
76	100	(defn tag-value
77		- "One IRCv3 tag's value, or nil. Tag values escape `;` and space; the ids and
78		- timestamps read here contain neither, so they are taken as they come."
	101	+ "One IRCv3 tag's value, unescaped, or nil."
79	102	[tags key]
80	103	(when tags
81	104	(some (fn [pair]
82	105	(let [[k v] (str/split pair #"=" 2)]
83		- (when (= k key) v)))
	106	+ (when (= k key) (unescape-tag v))))
84	107	(str/split tags #";"))))
85	108
86	109	(defn nick-of

       :command (str/upper-case (or (first parts) ""))
       :params (cond-> (vec (rest parts)) trailing (conj trailing))}))
++(defn unescape-tag
++  "An IRCv3 tag value with its escapes undone.
++
++  `\\:` is a semicolon, `\\s` a space, and `\\\\`, `\\r` and `\\n` themselves —
++  the escaping exists because `;` separates tags and a space ends them. It
++  matters for any value that can contain either: a reaction tally is
++  `emoji:nick;emoji:nick` on the wire and arrives with every one of those
++  semicolons written `\\:`, so a reader that skips this step sees one tally
++  where there were three, and counts to match."
++  [v]
++  (when v
++    (loop [in (seq v) out []]
++      (if-let [c (first in)]
++        (if (and (= \\ c) (second in))
++          (recur (drop 2 in)
++                 (conj out (case (second in)
++                             \: \;
++                             \s \space
++                             \r \return
++                             \n \newline
++                             (second in))))
++          (recur (rest in) (conj out c)))
++        (apply str out)))))
++
  (defn tag-value
--  "One IRCv3 tag's value, or nil. Tag values escape `;` and space; the ids and
++  "One IRCv3 tag's value, unescaped, or nil."
--  timestamps read here contain neither, so they are taken as they come."
    [tags key]
    (when tags
      (some (fn [pair]
              (let [[k v] (str/split pair #"=" 2)]
--              (when (= k key) v)))
++              (when (= k key) (unescape-tag v))))
            (str/split tags #";"))))
  (defn nick-of