fix #2844 #3911; add --spellsuggest to suggest symbols in scope with similar spellings on undefined symbol error (#16067)

* add --spellsuggest to suggest symbols in scope with similar spellings on undefined symbol errors
* implement --spellsuggest with 0 arguments

changelog.md

@@ -261,6 +261,8 @@
 
 - Deprecated `--nilseqs` which is now a noop.
 
+- Added `--spellSuggest` to show spelling suggestions on typos.
+
 - Source+Edit links now appear on top of every docgen'd page when
   `nim doc --git.url:url ...` is given.
 

compiler/commands.nim

@@ -863,6 +863,10 @@ proc processSwitch*(switch, arg: string, pass: TCmdLinePass, info: TLineInfo;
     processOnOffSwitchG(conf, {optStdout}, arg, pass, info)
   of "listfullpaths":
     processOnOffSwitchG(conf, {optListFullPaths}, arg, pass, info)
+  of "spellsuggest":
+    if arg.len == 0: conf.spellSuggestMax = spellSuggestSecretSauce
+    elif arg == "auto": conf.spellSuggestMax = spellSuggestSecretSauce
+    else: conf.spellSuggestMax = parseInt(arg)
   of "declaredlocs":
     processOnOffSwitchG(conf, {optDeclaredLocs}, arg, pass, info)
   of "dynliboverride":

compiler/lookups.nim

@@ -193,14 +193,17 @@ proc searchInScopes*(c: PContext, s: PIdent; ambiguous: var bool): PSym =
     if result != nil: return result
   result = someSymFromImportTable(c, s, ambiguous)
 
-proc debugScopes*(c: PContext; limit=0) {.deprecated.} =
+proc debugScopes*(c: PContext; limit=0, max = int.high) {.deprecated.} =
   var i = 0
+  var count = 0
   for scope in allScopes(c.currentScope):
     echo "scope ", i
     for h in 0..high(scope.symbols.data):
       if scope.symbols.data[h] != nil:
-        echo scope.symbols.data[h].name.s
-    if i == limit: break
+        if count >= max: return
+        echo count, ": ", scope.symbols.data[h].name.s
+        count.inc
+    if i == limit: return
     inc i
 
 proc searchInScopesFilterBy*(c: PContext, s: PIdent, filter: TSymKinds): seq[PSym] =
@@ -354,22 +357,60 @@ proc mergeShadowScope*(c: PContext) =
     else:
       c.addInterfaceDecl(sym)
 
-when defined(nimfix):
-  # when we cannot find the identifier, retry with a changed identifier:
-  proc altSpelling(x: PIdent): PIdent =
+when false:
+  # `nimfix` used to call `altSpelling` and prettybase.replaceDeprecated(n.info, ident, alt)
+  proc altSpelling(c: PContext, x: PIdent): PIdent =
     case x.s[0]
-    of 'A'..'Z': result = getIdent(toLowerAscii(x.s[0]) & x.s.substr(1))
-    of 'a'..'z': result = getIdent(toLowerAscii(x.s[0]) & x.s.substr(1))
+    of 'A'..'Z': result = getIdent(c.cache, toLowerAscii(x.s[0]) & x.s.substr(1))
+    of 'a'..'z': result = getIdent(c.cache, toLowerAscii(x.s[0]) & x.s.substr(1))
     else: result = x
 
-  template fixSpelling(n: PNode; ident: PIdent; op: untyped) =
-    let alt = ident.altSpelling
-    result = op(c, alt).skipAlias(n)
-    if result != nil:
-      prettybase.replaceDeprecated(n.info, ident, alt)
-      return result
-else:
-  template fixSpelling(n: PNode; ident: PIdent; op: untyped) = discard
+import std/[editdistance, heapqueue]
+
+type SpellCandidate = object
+  dist: int
+  depth: int
+  msg: string
+  sym: PSym
+
+template toOrderTup(a: SpellCandidate): auto =
+  # `dist` is first, to favor nearby matches
+  # `depth` is next, to favor nearby enclosing scopes among ties
+  # `sym.name.s` is last, to make the list ordered and deterministic among ties
+  (a.dist, a.depth, a.msg)
+
+proc `<`(a, b: SpellCandidate): bool =
+  a.toOrderTup < b.toOrderTup
+
+proc fixSpelling(c: PContext, n: PNode, ident: PIdent, result: var string) =
+  ## when we cannot find the identifier, suggest nearby spellings
+  if c.config.spellSuggestMax == 0: return
+  if c.compilesContextId > 0: return # don't slowdown inside compiles()
+  var list = initHeapQueue[SpellCandidate]()
+  let name0 = ident.s.nimIdentNormalize
+
+  for (sym, depth, isLocal) in allSyms(c):
+    let depth = -depth - 1
+    let dist = editDistance(name0, sym.name.s.nimIdentNormalize)
+    var msg: string
+    msg.add "\n ($1, $2): '$3'" % [$dist, $depth, sym.name.s]
+    addDeclaredLoc(msg, c.config, sym) # `msg` needed for deterministic ordering.
+    list.push SpellCandidate(dist: dist, depth: depth, msg: msg, sym: sym)
+
+  if list.len == 0: return
+  let e0 = list[0]
+  var count = 0
+  while true:
+    # pending https://github.com/timotheecour/Nim/issues/373 use more efficient `itemsSorted`.
+    if list.len == 0: break
+    let e = list.pop()
+    if c.config.spellSuggestMax == spellSuggestSecretSauce:
+      if e.dist > e0.dist: break
+    elif count >= c.config.spellSuggestMax: break
+    if count == 0:
+      result.add "\ncandidate misspellings (edit distance, lexical scope distance): "
+    result.add e.msg
+    count.inc
 
 proc errorUseQualifier(c: PContext; info: TLineInfo; s: PSym; amb: var bool): PSym =
   var err = "ambiguous identifier: '" & s.name.s & "'"
@@ -406,8 +447,8 @@ proc errorUseQualifier(c: PContext; info: TLineInfo; candidates: seq[PSym]) =
     inc i
   localError(c.config, info, errGenerated, err)
 
-proc errorUndeclaredIdentifier*(c: PContext; info: TLineInfo; name: string) =
-  var err = "undeclared identifier: '" & name & "'"
+proc errorUndeclaredIdentifier*(c: PContext; info: TLineInfo; name: string, extra = "") =
+  var err = "undeclared identifier: '" & name & "'" & extra
   if c.recursiveDep.len > 0:
     err.add "\nThis might be caused by a recursive module dependency:\n"
     err.add c.recursiveDep
@@ -415,25 +456,25 @@ proc errorUndeclaredIdentifier*(c: PContext; info: TLineInfo; name: string) =
     c.recursiveDep = ""
   localError(c.config, info, errGenerated, err)
 
+proc errorUndeclaredIdentifierHint*(c: PContext; n: PNode, ident: PIdent): PSym =
+  var extra = ""
+  fixSpelling(c, n, ident, extra)
+  errorUndeclaredIdentifier(c, n.info, ident.s, extra)
+  result = errorSym(c, n)
+
 proc lookUp*(c: PContext, n: PNode): PSym =
   # Looks up a symbol. Generates an error in case of nil.
   var amb = false
   case n.kind
   of nkIdent:
     result = searchInScopes(c, n.ident, amb).skipAlias(n, c.config)
-    if result == nil:
-      fixSpelling(n, n.ident, searchInScopes)
-      errorUndeclaredIdentifier(c, n.info, n.ident.s)
-      result = errorSym(c, n)
+    if result == nil: result = errorUndeclaredIdentifierHint(c, n, n.ident)
   of nkSym:
     result = n.sym
   of nkAccQuoted:
     var ident = considerQuotedIdent(c, n)
     result = searchInScopes(c, ident, amb).skipAlias(n, c.config)
-    if result == nil:
-      fixSpelling(n, ident, searchInScopes)
-      errorUndeclaredIdentifier(c, n.info, ident.s)
-      result = errorSym(c, n)
+    if result == nil: result = errorUndeclaredIdentifierHint(c, n, ident)
   else:
     internalError(c.config, n.info, "lookUp")
     return
@@ -471,9 +512,7 @@ proc qualifiedLookUp*(c: PContext, n: PNode, flags: set[TLookupFlag]): PSym =
           errorUseQualifier(c, n.info, candidates)
 
     if result == nil and checkUndeclared in flags:
-      fixSpelling(n, ident, searchInScopes)
-      errorUndeclaredIdentifier(c, n.info, ident.s)
-      result = errorSym(c, n)
+      result = errorUndeclaredIdentifierHint(c, n, ident)
     elif checkAmbiguity in flags and result != nil and amb:
       result = errorUseQualifier(c, n.info, result, amb)
     c.isAmbiguous = amb
@@ -494,9 +533,7 @@ proc qualifiedLookUp*(c: PContext, n: PNode, flags: set[TLookupFlag]): PSym =
         else:
           result = someSym(c.graph, m, ident).skipAlias(n, c.config)
         if result == nil and checkUndeclared in flags:
-          fixSpelling(n[1], ident, searchInScopes)
-          errorUndeclaredIdentifier(c, n[1].info, ident.s)
-          result = errorSym(c, n[1])
+          result = errorUndeclaredIdentifierHint(c, n[1], ident)
       elif n[1].kind == nkSym:
         result = n[1].sym
       elif checkUndeclared in flags and

compiler/options.nim

@@ -268,6 +268,7 @@ type
     numberOfProcessors*: int   # number of processors
     lastCmdTime*: float        # when caas is enabled, we measure each command
     symbolFiles*: SymbolFilesOption
+    spellSuggestMax*: int # max number of spelling suggestions for typos
 
     cppDefines*: HashSet[string] # (*)
     headerFile*: string
@@ -575,6 +576,7 @@ const
   htmldocsDir* = htmldocsDirname.RelativeDir
   docRootDefault* = "@default" # using `@` instead of `$` to avoid shell quoting complications
   oKeepVariableNames* = true
+  spellSuggestSecretSauce* = -1
 
 proc mainCommandArg*(conf: ConfigRef): string =
   ## This is intended for commands like check or parse

compiler/semcall.nim

@@ -407,6 +407,7 @@ proc resolveOverloads(c: PContext, n, orig: PNode,
 
     if overloadsState == csEmpty and result.state == csEmpty:
       if efNoUndeclared notin flags: # for tests/pragmas/tcustom_pragma.nim
+        # xxx adapt/use errorUndeclaredIdentifierHint(c, n, f.ident)
         localError(c.config, n.info, getMsgDiagnostic(c, flags, n, f))
       return
     elif result.state != csMatch:

doc/advopt.txt

@@ -35,6 +35,9 @@ Advanced options:
   --colors:on|off           turn compiler messages coloring on|off
   --listFullPaths:on|off    list full paths in messages
   --declaredLocs:on|off     show declaration locations in messages
+  --spellSuggest|:num       show at most `num >= 0` spelling suggestions on typos.
+                            if `num` is not specified (or `auto`), return
+                            an implementation defined set of suggestions.
   -w:on|off|list, --warnings:on|off|list
                             turn all warnings on|off or list all available
   --warning[X]:on|off       turn specific warning X on|off

tests/misc/mspellsuggest.nim

@@ -0,0 +1,7 @@
+proc fooBar4*(a: int) = discard
+var fooBar9* = 0
+
+var fooCar* = 0
+type FooBar* = int
+type FooCar* = int
+type GooBa* = int

tests/misc/tspellsuggest.nim

@@ -0,0 +1,45 @@
+discard """
+  # pending bug #16521 (bug 12) use `matrix`
+  cmd: "nim c --spellsuggest:15 --hints:off $file"
+  action: "reject"
+  nimout: '''
+tspellsuggest.nim(45, 13) Error: undeclared identifier: 'fooBar'
+candidate misspellings (edit distance, lexical scope distance):
+ (1, 0): 'fooBar8' [var declared in tspellsuggest.nim(43, 9)]
+ (1, 1): 'fooBar7' [var declared in tspellsuggest.nim(41, 7)]
+ (1, 3): 'fooBar1' [var declared in tspellsuggest.nim(33, 5)]
+ (1, 3): 'fooBar2' [let declared in tspellsuggest.nim(34, 5)]
+ (1, 3): 'fooBar3' [const declared in tspellsuggest.nim(35, 7)]
+ (1, 3): 'fooBar4' [proc declared in tspellsuggest.nim(36, 6)]
+ (1, 3): 'fooBar5' [template declared in tspellsuggest.nim(37, 10)]
+ (1, 3): 'fooBar6' [macro declared in tspellsuggest.nim(38, 7)]
+ (1, 5): 'FooBar' [type declared in mspellsuggest.nim(5, 6)]
+ (1, 5): 'fooBar4' [proc declared in mspellsuggest.nim(1, 6)]
+ (1, 5): 'fooBar9' [var declared in mspellsuggest.nim(2, 5)]
+ (1, 5): 'fooCar' [var declared in mspellsuggest.nim(4, 5)]
+ (2, 5): 'FooCar' [type declared in mspellsuggest.nim(6, 6)]
+ (2, 5): 'GooBa' [type declared in mspellsuggest.nim(7, 6)]
+ (3, 0): 'fooBarBaz' [const declared in tspellsuggest.nim(44, 11)]
+'''
+"""
+
+# tests `--spellsuggest:num`
+
+
+
+# line 30
+import ./mspellsuggest
+
+var fooBar1 = 0
+let fooBar2 = 0
+const fooBar3 = 0
+proc fooBar4() = discard
+template fooBar5() = discard
+macro fooBar6() = discard
+
+proc main =
+  var fooBar7 = 0
+  block:
+    var fooBar8 = 0
+    const fooBarBaz = 0
+    let x = fooBar

tests/misc/tspellsuggest2.nim

@@ -0,0 +1,45 @@
+discard """
+  # pending bug #16521 (bug 12) use `matrix`
+  cmd: "nim c --spellsuggest --hints:off $file"
+  action: "reject"
+  nimout: '''
+tspellsuggest2.nim(45, 13) Error: undeclared identifier: 'fooBar'
+candidate misspellings (edit distance, lexical scope distance):
+ (1, 0): 'fooBar8' [var declared in tspellsuggest2.nim(43, 9)]
+ (1, 1): 'fooBar7' [var declared in tspellsuggest2.nim(41, 7)]
+ (1, 3): 'fooBar1' [var declared in tspellsuggest2.nim(33, 5)]
+ (1, 3): 'fooBar2' [let declared in tspellsuggest2.nim(34, 5)]
+ (1, 3): 'fooBar3' [const declared in tspellsuggest2.nim(35, 7)]
+ (1, 3): 'fooBar4' [proc declared in tspellsuggest2.nim(36, 6)]
+ (1, 3): 'fooBar5' [template declared in tspellsuggest2.nim(37, 10)]
+ (1, 3): 'fooBar6' [macro declared in tspellsuggest2.nim(38, 7)]
+ (1, 5): 'FooBar' [type declared in mspellsuggest.nim(5, 6)]
+ (1, 5): 'fooBar4' [proc declared in mspellsuggest.nim(1, 6)]
+ (1, 5): 'fooBar9' [var declared in mspellsuggest.nim(2, 5)]
+ (1, 5): 'fooCar' [var declared in mspellsuggest.nim(4, 5)]
+'''
+"""
+
+# tests `--spellsuggest`
+
+
+
+
+
+
+# line 30
+import ./mspellsuggest
+
+var fooBar1 = 0
+let fooBar2 = 0
+const fooBar3 = 0
+proc fooBar4() = discard
+template fooBar5() = discard
+macro fooBar6() = discard
+
+proc main =
+  var fooBar7 = 0
+  block:
+    var fooBar8 = 0
+    const fooBarBaz = 0
+    let x = fooBar