Improve tweet url and hashtag parsing

src/parserutils.nim

@@ -1,4 +1,4 @@
-import strutils, times, macros, htmlgen, unicode, options
+import strutils, times, macros, htmlgen, unicode, options, algorithm
 import regex, packedjson
 import types, utils, formatters
 
@@ -6,9 +6,18 @@ const
   unRegex = re"(^|[^A-z0-9-_./?])@([A-z0-9_]{1,15})"
   unReplace = "$1<a href=\"/$2\">@$2</a>"
 
-  htRegex = re"(^|[^\w-_./?])([#$])([\w_]+)"
+  htRegex = re"(^|[^\w-_./?])([#＃$])([\w_]+)"
   htReplace = "$1<a href=\"/search?q=%23$3\">$2$3</a>"
 
+type
+  ReplaceSliceKind = enum
+    rkRemove, rkUrl, rkHashtag, rkMention
+
+  ReplaceSlice = object
+    slice: Slice[int]
+    kind: ReplaceSliceKind
+    url, display: string
+
 template isNull*(js: JsonNode): bool = js.kind == JNull
 template notNull*(js: JsonNode): bool = js.kind != JNull
 
@@ -124,65 +133,97 @@ proc getTombstone*(js: JsonNode): string =
   result = js{"tombstoneInfo", "richText", "text"}.getStr
   result.removeSuffix(" Learn more")
 
-template getSlice(text: string; slice: seq[int]): string =
+proc extractSlice(js: JsonNode): Slice[int] =
-  text.runeSubStr(slice[0], slice[1] - slice[0])
+  result = js["indices"][0].getInt ..< js["indices"][1].getInt
-
-proc getSlice(text: string; js: JsonNode): string =
-  if js.kind != JArray or js.len < 2 or js[0].kind != JInt: return text
-
-  let slice = @[js{0}.getInt, js{1}.getInt]
-  text.getSlice(slice)
-
-proc expandUrl(text: var string; js: JsonNode; tLen: int; hideTwitter=false) =
-  let u = js{"url"}.getStr
-  if u.len == 0 or u notin text:
-    return
 
+proc extractUrls(result: var seq[ReplaceSlice]; js: JsonNode;
+                 textLen: int; hideTwitter = false) =
   let
-    url = js{"expanded_url"}.getStr
+    url = js["expanded_url"].getStr
-    slice = js{"indices"}[1].getInt
+    slice = js.extractSlice
 
-  if hideTwitter and slice >= tLen and url.isTwitterUrl:
+  if hideTwitter and slice.b.succ >= textLen and url.isTwitterUrl:
-    text = text.replace(u, "")
+    if slice.a < textLen:
-    text.removeSuffix(' ')
+      result.add ReplaceSlice(kind: rkRemove, slice: slice)
-    text.removeSuffix('\n')
   else:
-    text = text.replace(u, a(shortLink(url), href=url))
+    result.add ReplaceSlice(kind: rkUrl, url: url,
+                            display: url.shortLink, slice: slice)
 
-proc expandMention(text: var string; orig: string; js: JsonNode) =
+proc extractHashtags(result: var seq[ReplaceSlice]; js: JsonNode) =
-  let
+  result.add ReplaceSlice(kind: rkHashtag, slice: js.extractSlice)
-    name = js{"name"}.getStr
+
-    href = '/' & js{"screen_name"}.getStr
+proc replacedWith(runes: seq[Rune]; repls: openArray[ReplaceSlice];
-    uname = orig.getSlice(js{"indices"})
+                  textSlice: Slice[int]): string =
-  text = text.replace(uname, a(uname, href=href, title=name))
+  template extractLowerBound(i: int; idx): int =
+    if i > 0: repls[idx].slice.b.succ else: textSlice.a
+
+  result = newStringOfCap(runes.len)
+
+  var upperBound = textSlice.b
+  for i, rep in repls:
+    result.add $runes[extractLowerBound(i, i - 1) ..< rep.slice.a]
+    case rep.kind
+    of rkHashtag:
+      let
+        name = $runes[rep.slice.a.succ .. rep.slice.b]
+        symbol = $runes[rep.slice.a]
+      result.add a(symbol & name, href = "/search?q=%23" & name)
+    of rkMention:
+      result.add a($runes[rep.slice], href = rep.url, title = rep.display)
+    of rkUrl:
+      result.add a(rep.display, href = rep.url)
+    of rkRemove:
+      discard
+
+  let rest = extractLowerBound(repls.len, ^1) ..< textSlice.b
+  if rest.a <= rest.b:
+    result.add $runes[rest]
+
+proc deduplicate(s: var seq[ReplaceSlice]) =
+  var
+    len = s.len
+    i = 0
+  while i < len:
+    var j = i + 1
+    while j < len:
+      if s[i].slice.a == s[j].slice.a:
+        s.del j
+        dec len
+      else:
+        inc j
+    inc i
+
+proc cmp(x, y: ReplaceSlice): int = cmp(x.slice.a, y.slice.b)
 
 proc expandProfileEntities*(profile: var Profile; js: JsonNode) =
   let
-    orig = profile.bio
+    orig = profile.bio.toRunes
     ent = ? js{"entities"}
 
   with urls, ent{"url", "urls"}:
     profile.website = urls[0]{"expanded_url"}.getStr
 
-  with urls, ent{"description", "urls"}:
+  var replacements = newSeq[ReplaceSlice]()
-    for u in urls: profile.bio.expandUrl(u, orig.high)
 
+  with urls, ent{"description", "urls"}:
+    for u in urls:
+      replacements.extractUrls(u, orig.high)
+
+  replacements.deduplicate
+  replacements.sort(cmp)
+
+  profile.bio = orig.replacedWith(replacements, 0 .. orig.len)
   profile.bio = profile.bio.replace(unRegex, unReplace)
                            .replace(htRegex, htReplace)
 
-  for mention in ? ent{"user_mentions"}:
-    profile.bio.expandMention(orig, mention)
-
 proc expandTweetEntities*(tweet: Tweet; js: JsonNode) =
   let
-    orig = tweet.text
+    orig = tweet.text.toRunes
     textRange = js{"display_text_range"}
-    slice = @[textRange{0}.getInt, textRange{1}.getInt]
+    textSlice = textRange{0}.getInt .. textRange{1}.getInt
     hasQuote = js{"is_quote_status"}.getBool
     hasCard = tweet.card.isSome
 
-  tweet.text = tweet.text.getSlice(slice)
-
   var replyTo = ""
   if tweet.replyId != 0:
     with reply, js{"in_reply_to_screen_name"}:
@@ -191,26 +232,45 @@ proc expandTweetEntities*(tweet: Tweet; js: JsonNode) =
 
   let ent = ? js{"entities"}
 
+  var replacements = newSeq[ReplaceSlice]()
+
   with urls, ent{"urls"}:
     for u in urls:
-      tweet.text.expandUrl(u, slice[1], hasQuote)
+      let urlStr = u["url"].getStr
+      if urlStr.len == 0 or urlStr notin tweet.text:
+        continue
+      replacements.extractUrls(u, textSlice.b, hideTwitter = hasQuote)
       if hasCard and u{"url"}.getStr == get(tweet.card).url:
         get(tweet.card).url = u{"expanded_url"}.getStr
 
   with media, ent{"media"}:
-    for m in media: tweet.text.expandUrl(m, slice[1], hideTwitter=true)
+    for m in media:
+      replacements.extractUrls(m, textSlice.b, hideTwitter = true)
 
-  if "hashtags" in ent or "symbols" in ent:
+  if "hashtags" in ent:
-    tweet.text = tweet.text.replace(htRegex, htReplace)
+    for hashtag in ent["hashtags"]:
+      replacements.extractHashtags(hashtag)
 
-  for mention in ? ent{"user_mentions"}:
+  if "symbols" in ent:
-    let
+    for symbol in ent["symbols"]:
-      name = mention{"screen_name"}.getStr
+      replacements.extractHashtags(symbol)
-      idx = tweet.reply.find(name)
 
-    if mention{"indices"}[0].getInt >= slice[0]:
+  if "user_mentions" in ent:
-      tweet.text.expandMention(orig, mention)
+    for mention in ent["user_mentions"]:
-      if idx > -1 and name != replyTo:
+      let
-        tweet.reply.delete idx
+        name = mention{"screen_name"}.getStr
-    elif idx == -1 and tweet.replyId != 0:
+        slice = mention.extractSlice
-      tweet.reply.add name
+        idx = tweet.reply.find(name)
+
+      if slice.a >= textSlice.a:
+        replacements.add ReplaceSlice(kind: rkMention, slice: slice,
+          url: "/" & name, display: mention["name"].getStr)
+        if idx > -1 and name != replyTo:
+          tweet.reply.delete idx
+      elif idx == -1 and tweet.replyId != 0:
+        tweet.reply.add name
+
+  replacements.deduplicate
+  replacements.sort(cmp)
+
+  tweet.text = orig.replacedWith(replacements, textSlice)