Fix OptTable::findNearest() adding delimiter for free

[thakis]

Summary

Prior to this, OptTable::findNearest() thought that the input --foo had an editing distance of 0 from an existing flag --foo=, which made it suggest flags with delimiters more often than flags without one. After this, it correctly assigns this case an editing distance of 1.

[MaskRay]

llvm/lib/Option/OptTable.cpp

294 ↗

(On Diff #197531)

How about:

StringRef LHS, RHS;
char Last = CandidateName.back();
std::string NormalizedName = Option;
if (Last == '=' || Last == ':') {
  std::tie(LHS, RHS) = Option.split(Last);
  NormalizedName = LHS;
  if (Option.find(Last) == LHS.size())
    NormalizedName += Last;
}

llvm/unittests/Option/OptionParsingTest.cpp

301 ↗

(On Diff #197531)

// --glormp should have an editing distance of 1 to --glormp=.

[thakis]

address comments

[thakis]

Thanks for the fast review!

llvm/lib/Option/OptTable.cpp

294 ↗

(On Diff #197531)

I like it.

But in a follow-up, I wanted to add the following the the loop below:

if (RHS.empty() && !Delimiter.empty()) {
  // The Candidate ends with a = or : delimiter, but the option passed in
  // didn't contain the delimiter (or doesn't have anything after it).
  // In that case, penalize the correction: `-nodefaultlibs` is more
  // likely to be a spello for `-nodefaultlib` than `-nodefaultlib:`
  // since the latter would need an argument, even though both have
  // an unmodified editing distance of 1.
  Distance++;
}

With Delimiter gone, that would become if (RHS.empty() && (Last == '=' || Last == ':') -- so I added a CandidateHasDelimiter variable too.

llvm/unittests/Option/OptionParsingTest.cpp

301 ↗

(On Diff #197531)

Whoops! Thanks.

[MaskRay]

Still looks good.

llvm/trunk/lib/Option/OptTable.cpp

for (const Info &CandidateInfo :
     // * Ignore positional argument option candidates (which do not
     //   have prefixes).
     if (!CandidateInfo.Prefixes)
       continue;
 
     // Now check if the candidate ends with a character commonly used when
     // delimiting an option from its value, such as '=' or ':'. If it does,
     // attempt to split the given option based on that delimiter.
-    std::string Delimiter = "";
-    char Last = CandidateName.back();
-    if (Last == '=' || Last == ':')
-      Delimiter = std::string(1, Last);
-
     StringRef LHS, RHS;
-    if (Delimiter.empty())
-      LHS = Option;
-    else
+    char Last = CandidateName.back();
+    bool CandidateHasDelimiter = Last == '=' || Last == ':';
+    std::string NormalizedName = Option;
+    if (CandidateHasDelimiter) {
       std::tie(LHS, RHS) = Option.split(Last);
+      NormalizedName = LHS;
+      if (Option.find(Last) == LHS.size())
+        NormalizedName += Last;
+    }
 
     // Consider each possible prefix for each candidate to find the most
     // appropriate one. For example, if a user asks for "--helm", suggest
     // "--help" over "-help".
     for (int P = 0;
          const char *const CandidatePrefix = CandidateInfo.Prefixes[P]; P++) {
-      std::string NormalizedName = (LHS + Delimiter).str();
       std::string Candidate = (CandidatePrefix + CandidateName).str();
       StringRef CandidateRef = Candidate;
       unsigned Distance =
           CandidateRef.edit_distance(NormalizedName, /*AllowReplacements=*/true,
                                      /*MaxEditDistance=*/BestDistance);
       if (Distance < BestDistance) {
         BestDistance = Distance;
         NearestString = (Candidate + RHS).str();

llvm/trunk/unittests/Option/OptionParsingTest.cpp

TEST(Option, FindNearest) {
   EXPECT_EQ(1U, T.findNearest("-fjormp", Nearest));
   EXPECT_EQ(Nearest, "--fjormp");
 
   // The nearest candidate respects the prefix and value delimiter
   // of the original string.
   EXPECT_EQ(1U, T.findNearest("/framb:foo", Nearest));
   EXPECT_EQ(Nearest, "/cramb:foo");
 
+  // `--glormp` should have an editing distance of 1 to `--glormp=`.
+  EXPECT_EQ(1U, T.findNearest("--glormp", Nearest));
+  EXPECT_EQ(Nearest, "--glormp=");
+  EXPECT_EQ(0U, T.findNearest("--glormp=foo", Nearest));
+
   // Flags should be included and excluded as specified.
   EXPECT_EQ(1U, T.findNearest("-doopf", Nearest, /*FlagsToInclude=*/OptFlag2));
   EXPECT_EQ(Nearest, "-doopf2");
   EXPECT_EQ(1U, T.findNearest("-doopf", Nearest,
                               /*FlagsToInclude=*/0,
                               /*FlagsToExclude=*/OptFlag2));
   EXPECT_EQ(Nearest, "-doopf1");
 }

llvm/trunk/unittests/Option/Opts.td

 
 def Blorp : Flag<["-", "--"], "blorp">, HelpText<"The blorp option">, Flags<[OptFlag1]>;
 def Blarn : Flag<["--", "-"], "blarn">, HelpText<"The blarn option">, Flags<[OptFlag1]>;
 def Cramb : Joined<["/"], "cramb:">, HelpText<"The cramb option">, MetaVarName<"CRAMB">, Flags<[OptFlag1]>;
 def Doopf1 : Flag<["-"], "doopf1">, HelpText<"The doopf1 option">, Flags<[OptFlag1]>;
 def Doopf2 : Flag<["-"], "doopf2">, HelpText<"The doopf2 option">, Flags<[OptFlag2]>;
 def Ermgh : Joined<["--"], "ermgh">, HelpText<"The ermgh option">, MetaVarName<"ERMGH">, Flags<[OptFlag1]>;
 def Fjormp : Flag<["--"], "fjormp">, HelpText<"The fjormp option">, Flags<[OptFlag1]>;
+
+def Glormp_eq : Flag<["--"], "glormp=">;
+
 def DashDash : Option<["--"], "", KIND_REMAINING_ARGS>;