regex: re-work compileSearchRegex

I've gotten tired of not being able to search for forward slashes.

Now it works like in vim, and you can also set default ignore case in
the config.

9 files changed, 127 insertions, 79 deletions

diff --git a/src/bindings/libregexp.nim b/src/bindings/libregexp.nim
index 85e7c1ca..d4c10b42 100644
--- a/src/bindings/libregexp.nim
+++ b/src/bindings/libregexp.nim
@@ -1,10 +1,19 @@
-const
-  LRE_FLAG_GLOBAL* = 1 shl 0
-  LRE_FLAG_IGNORECASE* = 1 shl 1
-  LRE_FLAG_MULTILINE* = 1 shl 2
-  LRE_FLAG_DOTALL* = 1 shl 3
-  LRE_FLAG_UTF16* = 1 shl 4
-  LRE_FLAG_STICKY* = 1 shl 5
+type
+  LREFlag* {.size: sizeof(cint).} = enum
+    LRE_FLAG_GLOBAL = "g"
+    LRE_FLAG_IGNORECASE = "i"
+    LRE_FLAG_MULTILINE = "m"
+    LRE_FLAG_DOTALL = "s"
+    LRE_FLAG_UTF16 = "u"
+    LRE_FLAG_STICKY = "y"
+
+  LREFlags* = set[LREFlag]
+
+func toCInt*(flags: LREFlags): cint =
+  cast[cint](flags)
+
+func toLREFlags*(flags: cint): LREFlags =
+  cast[LREFlags](flags)
 
 {.passc: "-Ilib/".}
 
diff --git a/src/config/config.nim b/src/config/config.nim
index 678f7f7b..7be0c4a2 100644
--- a/src/config/config.nim
+++ b/src/config/config.nim
@@ -81,6 +81,7 @@ type
 
   SearchConfig = object
     wrap* {.jsgetset.}: bool
+    default_flags* {.jsgetset.}: LREFlags
 
   EncodingConfig = object
     display_charset* {.jsgetset.}: Opt[Charset]
@@ -107,8 +108,8 @@ type
 
   DisplayConfig = object
     color_mode* {.jsgetset.}: Opt[ColorMode]
-    format_mode*: Opt[FormatMode] #TODO getset
-    no_format_mode*: FormatMode #TODO getset
+    format_mode* {.jsgetset.}: Opt[FormatMode]
+    no_format_mode* {.jsgetset.}: FormatMode
     emulate_overline* {.jsgetset.}: bool
     alt_screen* {.jsgetset.}: Opt[bool]
     highlight_color* {.jsgetset.}: RGBAColor
@@ -470,6 +471,7 @@ proc parseConfigValue[T](x: var Opt[T], v: TomlValue, k: string)
 proc parseConfigValue(x: var ActionMap, v: TomlValue, k: string)
 proc parseConfigValue(x: var CSSConfig, v: TomlValue, k: string)
 proc parseConfigValue[U, V](x: var Table[U, V], v: TomlValue, k: string)
+proc parseConfigValue[T](x: var set[T], v: TomlValue, k: string)
 
 proc typeCheck(v: TomlValue, vt: ValueType, k: string) =
   if v.vt != vt:
@@ -566,9 +568,10 @@ proc parseConfigValue(x: var Opt[FormatMode], v: TomlValue, k: string) =
 proc parseConfigValue(x: var FormatMode, v: TomlValue, k: string) =
   typeCheck(v, VALUE_ARRAY, k)
   for i in 0 ..< v.a.len:
-    let s = v.a[i].s
     let kk = k & "[" & $i & "]"
-    case s
+    let vv = v.a[i]
+    typeCheck(vv, VALUE_STRING, kk)
+    case vv.s
     of "bold": x.incl(FLAG_BOLD)
     of "italic": x.incl(FLAG_ITALIC)
     of "underline": x.incl(FLAG_UNDERLINE)
@@ -577,7 +580,7 @@ proc parseConfigValue(x: var FormatMode, v: TomlValue, k: string) =
     of "overline": x.incl(FLAG_OVERLINE)
     of "blink": x.incl(FLAG_BLINK)
     else:
-      raise newException(ValueError, "unknown format mode '" & s &
+      raise newException(ValueError, "unknown format mode '" & vv.s &
         "' for key " & kk)
 
 proc parseConfigValue(x: var RGBAColor, v: TomlValue, k: string) =
@@ -615,6 +618,27 @@ proc parseConfigValue(x: var ActionMap, v: TomlValue, k: string) =
       discard x.hasKeyOrPut(buf, "client.feedNext()")
     x[rk] = vv.s
 
+proc parseConfigValue[T: enum](x: var T, v: TomlValue, k: string) =
+  typeCheck(v, VALUE_STRING, k)
+  let e = strictParseEnum[T](v.s)
+  if e.isNone:
+    raise newException(ValueError, "invalid value '" & v.s & "' for key " & k)
+  x = e.get
+
+proc parseConfigValue[T](x: var set[T], v: TomlValue, k: string) =
+  typeCheck(v, {VALUE_STRING, VALUE_ARRAY}, k)
+  if v.vt == VALUE_STRING:
+    var xx: T
+    xx.parseConfigValue(v, k)
+    x = {xx}
+  else:
+    x = {}
+    for i in 0 ..< v.a.len:
+      let kk = k & "[" & $i & "]"
+      var xx: T
+      xx.parseConfigValue(v.a[i], kk)
+      x.incl(xx)
+
 var gdir {.compileTime.}: string
 proc parseConfigValue(x: var CSSConfig, v: TomlValue, k: string) =
   typeCheck(v, VALUE_TABLE, k)
diff --git a/src/js/fromjs.nim b/src/js/fromjs.nim
index 823a045e..f0ff8a91 100644
--- a/src/js/fromjs.nim
+++ b/src/js/fromjs.nim
@@ -10,6 +10,7 @@ import js/jstypes
 import js/opaque
 import js/tojs
 import types/opt
+import utils/twtstr
 
 proc fromJS*[T](ctx: JSContext, val: JSValue): JSResult[T]
 
@@ -230,7 +231,7 @@ proc fromJSSeq[T](ctx: JSContext, val: JSValue): JSResult[seq[T]] =
     s.add(genericRes.get)
   return ok(s)
 
-proc fromJSSet[T](ctx: JSContext, val: JSValue): Opt[set[T]] =
+proc fromJSSet[T](ctx: JSContext, val: JSValue): JSResult[set[T]] =
   let itprop = JS_GetProperty(ctx, val, ctx.getOpaque().sym_refs[ITERATOR])
   if JS_IsException(itprop):
     return err()
@@ -249,14 +250,14 @@ proc fromJSSet[T](ctx: JSContext, val: JSValue): Opt[set[T]] =
     if JS_IsException(next):
       return err()
     defer: JS_FreeValue(ctx, next)
-    let doneVal = JS_GetProperty(ctx, next, ctx.getOpaque().done)
+    let doneVal = JS_GetProperty(ctx, next, ctx.getOpaque().str_refs[DONE])
     if JS_IsException(doneVal):
       return err()
     defer: JS_FreeValue(ctx, doneVal)
     let done = ?fromJS[bool](ctx, doneVal)
     if done:
       break
-    let valueVal = JS_GetProperty(ctx, next, ctx.getOpaque().value)
+    let valueVal = JS_GetProperty(ctx, next, ctx.getOpaque().str_refs[VALUE])
     if JS_IsException(valueVal):
       return err()
     defer: JS_FreeValue(ctx, valueVal)
@@ -377,19 +378,9 @@ proc fromJSEnum[T: enum](ctx: JSContext, val: JSValue): JSResult[T] =
   if JS_IsException(val):
     return err()
   let s = ?toString(ctx, val)
-  # cmp when len is small enough, otherwise hashmap
-  when {T.low..T.high}.len <= 4:
-    for e in T.low .. T.high:
-      if $e == s:
-        return ok(e)
-  else:
-    const tab = (func(): Table[string, T] =
-      result = initTable[string, T]()
-      for e in T.low .. T.high:
-        result[$e] = e
-    )()
-    if s in tab:
-      return ok(tab[s])
+  let r = strictParseEnum[T](s)
+  if r.isSome:
+    return ok(r.get)
   return errTypeError("`" & s & "' is not a valid value for enumeration " & $T)
 
 proc fromJSPObj0(ctx: JSContext, val: JSValue, t: string):
diff --git a/src/js/javascript.nim b/src/js/javascript.nim
index 9e814386..d208ffad 100644
--- a/src/js/javascript.nim
+++ b/src/js/javascript.nim
@@ -170,6 +170,7 @@ proc free*(ctx: var JSContext) =
     JS_FreeValue(ctx, opaque.Array_prototype_values)
     JS_FreeValue(ctx, opaque.Object_prototype_valueOf)
     JS_FreeValue(ctx, opaque.Uint8Array_ctor)
+    JS_FreeValue(ctx, opaque.Set_ctor)
     for v in opaque.err_ctors:
       JS_FreeValue(ctx, v)
     GC_unref(opaque)
diff --git a/src/js/opaque.nim b/src/js/opaque.nim
index 135b6994..bae2bd5a 100644
--- a/src/js/opaque.nim
+++ b/src/js/opaque.nim
@@ -33,6 +33,7 @@ type
     Array_prototype_values*: JSValue
     Object_prototype_valueOf*: JSValue
     Uint8Array_ctor*: JSValue
+    Set_ctor*: JSValue
     err_ctors*: array[JSErrorEnum, JSValue]
     htmldda*: JSClassID # only one of these exists: document.all.
 
@@ -71,8 +72,11 @@ func newJSContextOpaque*(ctx: JSContext): JSContextOpaque =
       opaque.Object_prototype_valueOf = JS_GetPropertyStr(ctx, objproto, "valueOf")
       JS_FreeValue(ctx, objproto)
     block:
-      let u8actor = JS_GetPropertyStr(ctx, global, "Uint8Array")
-      opaque.Uint8Array_ctor = u8actor
+      opaque.Set_ctor = JS_GetPropertyStr(ctx, global, "Set")
+      assert not JS_IsException(opaque.Set_ctor)
+    block:
+      opaque.Uint8Array_ctor = JS_GetPropertyStr(ctx, global, "Uint8Array")
+      assert not JS_IsException(opaque.Uint8Array_ctor)
     for e in JSErrorEnum:
       let s = $e
       let err = JS_GetPropertyStr(ctx, global, cstring(s))
diff --git a/src/js/regex.nim b/src/js/regex.nim
index fdc9e8e0..1d78f806 100644
--- a/src/js/regex.nim
+++ b/src/js/regex.nim
@@ -6,13 +6,7 @@ import bindings/quickjs
 import types/opt
 import utils/twtstr
 
-export
-  LRE_FLAG_GLOBAL,
-  LRE_FLAG_IGNORECASE,
-  LRE_FLAG_MULTILINE,
-  LRE_FLAG_DOTALL,
-  LRE_FLAG_UTF16,
-  LRE_FLAG_STICKY
+export LREFlags
 
 type
   Regex* = object
@@ -34,13 +28,13 @@ var dummyContext = JS_NewContextRaw(dummyRuntime)
 func `$`*(regex: Regex): string =
   regex.buf
 
-proc compileRegex*(buf: string, flags: int): Result[Regex, string] =
+proc compileRegex*(buf: string, flags: LREFlags = {}): Result[Regex, string] =
   var error_msg_size = 64
   var error_msg = newString(error_msg_size)
   prepareMutation(error_msg)
   var plen: cint
   let bytecode = lre_compile(addr plen, cstring(error_msg),
-    cint(error_msg_size), cstring(buf), csize_t(buf.len), cint(flags),
+    cint(error_msg_size), cstring(buf), csize_t(buf.len), flags.toCInt,
     dummyContext)
   if bytecode == nil:
     return err(error_msg.until('\0')) # Failed to compile.
@@ -68,16 +62,16 @@ func countBackslashes(buf: string, i: int): int =
 # mnop -> ^mnop$
 proc compileMatchRegex*(buf: string): Result[Regex, string] =
   if buf.len == 0:
-    return compileRegex(buf, 0)
+    return compileRegex(buf)
   if buf[0] == '^':
-    return compileRegex(buf, 0)
+    return compileRegex(buf)
   if buf[^1] == '$':
     # Check whether the final dollar sign is escaped.
     if buf.len == 1 or buf[^2] != '\\':
-      return compileRegex(buf, 0)
+      return compileRegex(buf)
     let j = buf.countBackslashes(buf.high - 2)
     if j mod 2 == 1: # odd, because we do not count the last backslash
-      return compileRegex(buf, 0)
+      return compileRegex(buf)
     # escaped. proceed as if no dollar sign was at the end
   if buf[^1] == '\\':
     # Check if the regex contains an invalid trailing backslash.
@@ -87,38 +81,34 @@ proc compileMatchRegex*(buf: string): Result[Regex, string] =
   var buf2 = "^"
   buf2 &= buf
   buf2 &= "$"
-  return compileRegex(buf2, 0)
-
-proc compileSearchRegex*(str: string): Result[Regex, string] =
-  # Parse any applicable flags in regex/<flags>. The last forward slash is
-  # dropped when <flags> is empty, and interpreted as a character when the
-  # flags are is invalid.
-
-  var i = str.high
-  var flagsi = -1
-  while i >= 0:
-    case str[i]
-    of '/':
-      flagsi = i
-      break
-    of 'i', 'm', 's', 'u': discard
-    else: break # invalid flag
-    dec i
-
-  var flags = LRE_FLAG_GLOBAL # for easy backwards matching
-
-  if flagsi == -1:
-    return compileRegex(str, flags)
-
-  for i in flagsi..str.high:
-    case str[i]
-    of '/': discard
-    of 'i': flags = flags or LRE_FLAG_IGNORECASE
-    of 'm': flags = flags or LRE_FLAG_MULTILINE
-    of 's': flags = flags or LRE_FLAG_DOTALL
-    of 'u': flags = flags or LRE_FLAG_UTF16
-    else: assert false
-  return compileRegex(str.substr(0, flagsi - 1), flags)
+  return compileRegex(buf2)
+
+proc compileSearchRegex*(str: string, defaultFlags: LREFlags):
+    Result[Regex, string] =
+  # Emulate vim's \c/\C: override defaultFlags if one is found, then remove it
+  # from str.
+  var flags = defaultFlags
+  var s = newStringOfCap(str.len)
+  var quot = false
+  for c in str:
+    if quot:
+      quot = false
+      if c == 'c':
+        flags.incl(LRE_FLAG_IGNORECASE)
+        continue
+      elif c == 'C':
+        flags.excl(LRE_FLAG_IGNORECASE)
+        continue
+      else:
+        s &= '\\'
+    if c == '\\':
+      quot = true
+    else:
+      s &= c
+  if quot:
+    s &= '\\'
+  flags.incl(LRE_FLAG_GLOBAL) # for easy backwards matching
+  return compileRegex(s, flags)
 
 proc exec*(regex: Regex, str: string, start = 0, length = -1, nocaps = false): RegexResult =
   let length = if length == -1:
@@ -134,7 +124,7 @@ proc exec*(regex: Regex, str: string, start = 0, length = -1, nocaps = false): R
     let size = sizeof(ptr uint8) * captureCount * 2
     capture = cast[ptr UncheckedArray[int]](alloc0(size))
   var cstr = cstring(str)
-  let flags = lre_get_flags(bytecode)
+  let flags = lre_get_flags(bytecode).toLREFlags
   var start = start
   while true:
     let ret = lre_exec(cast[ptr ptr uint8](capture), bytecode,
@@ -151,7 +141,7 @@ proc exec*(regex: Regex, str: string, start = 0, length = -1, nocaps = false): R
       let s = capture[i * 2] - cstrAddress
       let e = capture[i * 2 + 1] - cstrAddress
       result.captures.add((s, e))
-    if (flags and LRE_FLAG_GLOBAL) != 1:
+    if LRE_FLAG_GLOBAL notin flags:
       break
     if start >= str.len:
       break
diff --git a/src/js/tojs.nim b/src/js/tojs.nim
index c79d1bf2..757552eb 100644
--- a/src/js/tojs.nim
+++ b/src/js/tojs.nim
@@ -64,6 +64,7 @@ proc toJS*[U, V](ctx: JSContext, t: Table[U, V]): JSValue
 proc toJS*(ctx: JSContext, opt: Option): JSValue
 proc toJS*[T, E](ctx: JSContext, opt: Result[T, E]): JSValue
 proc toJS*(ctx: JSContext, s: seq): JSValue
+proc toJS*[T](ctx: JSContext, s: set[T]): JSValue
 proc toJS*(ctx: JSContext, t: tuple): JSValue
 proc toJS*(ctx: JSContext, e: enum): JSValue
 proc toJS*(ctx: JSContext, j: JSValue): JSValue
@@ -209,6 +210,16 @@ proc toJS(ctx: JSContext, s: seq): JSValue =
         return JS_EXCEPTION
   return a
 
+proc toJS*[T](ctx: JSContext, s: set[T]): JSValue =
+  #TODO this is a bit lazy :p
+  var x = newSeq[T]()
+  for e in s:
+    x.add(e)
+  var a = toJS(ctx, x)
+  if JS_IsException(a):
+    return a
+  return JS_CallConstructor(ctx, ctx.getOpaque().Set_ctor, 1, addr a)
+
 proc toJS(ctx: JSContext, t: tuple): JSValue =
   let a = JS_NewArray(ctx)
   if not JS_IsException(a):
diff --git a/src/local/pager.nim b/src/local/pager.nim
index aa546082..d30a0dbc 100644
--- a/src/local/pager.nim
+++ b/src/local/pager.nim
@@ -807,7 +807,8 @@ proc updateReadLineISearch(pager: Pager, linemode: LineMode) =
       pager.isearchpromise = nil
     of EDIT:
       if lineedit.news != "":
-        pager.iregex = compileSearchRegex(lineedit.news)
+        pager.iregex = compileSearchRegex(lineedit.news,
+          pager.config.search.default_flags)
       pager.container.popCursorPos(true)
       pager.container.pushCursorPos()
       if pager.iregex.isSome:
@@ -857,7 +858,8 @@ proc updateReadLine*(pager: Pager) =
       of BUFFER: pager.container.readSuccess(lineedit.news)
       of SEARCH_F, SEARCH_B:
         if lineedit.news != "":
-          pager.regex = pager.checkRegex(compileSearchRegex(lineedit.news))
+          pager.regex = pager.checkRegex(compileSearchRegex(lineedit.news,
+            pager.config.search.default_flags))
         pager.reverseSearch = pager.linemode == SEARCH_B
         pager.searchNext()
       of GOTO_LINE:
diff --git a/src/utils/twtstr.nim b/src/utils/twtstr.nim
index 80b6be65..1916fff5 100644
--- a/src/utils/twtstr.nim
+++ b/src/utils/twtstr.nim
@@ -713,3 +713,19 @@ proc makeCRLF*(s: string): string =
       result &= '\n'
     else:
       result &= s[i]
+
+func strictParseEnum*[T: enum](s: string): Opt[T] =
+  # cmp when len is small enough, otherwise hashmap
+  when {T.low..T.high}.len <= 4:
+    for e in T.low .. T.high:
+      if $e == s:
+        return ok(e)
+  else:
+    const tab = (func(): Table[string, T] =
+      result = initTable[string, T]()
+      for e in T.low .. T.high:
+        result[$e] = e
+    )()
+    if s in tab:
+      return ok(tab[s])
+  return err()