[TableGen] Eliminate the 'code' type

Update the documentation.

Rework various backends that relied on the code type.

Differential Revision: https://reviews.llvm.org/D92269

clang/utils/TableGen/ClangOptionDocEmitter.cpp

@@ -217,8 +217,6 @@ std::string getRSTStringWithTextFallback(const Record *R, StringRef Primary,
       StringRef Value;
       if (auto *SV = dyn_cast_or_null<StringInit>(V->getValue()))
         Value = SV->getValue();
-      else if (auto *CV = dyn_cast_or_null<CodeInit>(V->getValue()))
-        Value = CV->getValue();
       if (!Value.empty())
         return Field == Primary ? Value.str() : escapeRST(Value);
     }

llvm/docs/TableGen/BackEnds.rst

@@ -693,8 +693,8 @@ This class provides six fields.
   table that holds the entries. If unspecified, the ``FilterClass`` name is
   used.
 
-* ``list<string> Fields``. A list of the names of the fields in the
-  collected records that contain the data for the table entries. The order of
+* ``list<string> Fields``. A list of the names of the fields *in the
+  collected records* that contain the data for the table entries. The order of
   this list determines the order of the values in the C++ initializers. See
   below for information about the types of these fields.
 
@@ -706,13 +706,26 @@ This class provides six fields.
 
 * ``bit PrimaryKeyEarlyOut``. See the third example below.
 
-TableGen attempts to deduce the type of each of the table fields. It can
-deduce ``bit``, ``bits<n>``, ``string``, ``Intrinsic``, and ``Instruction``.
-These can be used in the primary key. TableGen also deduces ``code``, but it
-cannot be used in the primary key. Any other field types must be specified
+TableGen attempts to deduce the type of each of the table fields so that it
+can format the C++ initializers in the emitted table. It can deduce ``bit``,
+``bits<n>``, ``string``, ``Intrinsic``, and ``Instruction``.  These can be
+used in the primary key. Any other field types must be specified
 explicitly; this is done as shown in the second example below. Such fields
 cannot be used in the primary key.
 
+One special case of the field type has to do with code. Arbitrary code is
+represented by a string, but has to be emitted as a C++ initializer without
+quotes. If the code field was defined using a code literal (``[{...}]``),
+then TableGen will know to emit it without quotes. However, if it was
+defined using a string literal or complex string expression, then TableGen
+will not know. In this case, you can force TableGen to treat the field as
+code by including the following line in the ``GenericTable`` record, where
+*xxx* is the code field name.
+
+.. code-block:: text
+
+  string TypeOf_xxx = "code";
+
 Here is an example where TableGen can deduce the field types. Note that the
 table entry records are anonymous; the names of entry records are
 irrelevant.
@@ -793,7 +806,7 @@ pointer if no entry is found.
 
 This example includes a field whose type TableGen cannot deduce. The ``Kind``
 field uses the enumerated type ``CEnum`` defined above. To inform TableGen
-of the type, the class derived from ``GenericTable`` must include a field
+of the type, the record derived from ``GenericTable`` must include a string field
 named ``TypeOf_``\ *field*, where *field* is the name of the field whose type
 is required.
 
@@ -802,7 +815,7 @@ is required.
   def CTable : GenericTable {
     let FilterClass = "CEntry";
     let Fields = ["Name", "Kind", "Encoding"];
-    GenericEnum TypeOf_Kind = CEnum;
+    string TypeOf_Kind = "CEnum";
     let PrimaryKey = ["Encoding"];
     let PrimaryKeyName = "lookupCEntryByEncoding";
   }

llvm/docs/TableGen/BackGuide.rst

@@ -287,9 +287,9 @@ value. The static function ``get()`` can be used to obtain the singleton
 
 This class, a subclass of ``Init``, acts as the parent class of the classes
 that represent specific value types (except for the unset value). These
-classes include ``BitInit``, ``BitsInit``, ``CodeInit``, ``DagInit``,
-``DefInit``, ``IntInit``, ``ListInit``, and ``StringInit``. (There are
-additional derived types used by the TableGen parser.)
+classes include ``BitInit``, ``BitsInit``, ``DagInit``, ``DefInit``,
+``IntInit``, ``ListInit``, and ``StringInit``. (There are additional derived
+types used by the TableGen parser.)
 
 This class includes a data member that specifies the ``RecTy`` type of the
 value. It provides a function to get that ``RecTy`` type.
@@ -330,18 +330,6 @@ The class provides the following additional functions.
 
 * A function that gets a bit specified by an integer index.
 
-``CodeInit``
-~~~~~~~~~~~~
-
-The ``CodeInit`` class is a subclass of ``TypedInit``. Its instances
-represent arbitrary-length strings produced from ``code`` literals in the
-TableGen files. It includes a data member that contains a ``StringRef`` of
-the value.
-
-The class provides the usual ``get()`` and ``getValue()`` functions. The
-latter function returns the ``StringRef``.
-
-
 ``DagInit``
 ~~~~~~~~~~~
 

llvm/docs/TableGen/ProgRef.rst

@@ -167,10 +167,11 @@ TableGen has two kinds of string literals:
 
 .. productionlist::
    TokString: '"' (non-'"' characters and escapes) '"'
-   TokCodeFragment: "[{" (shortest text not containing "}]") "}]"
+   TokCode: "[{" (shortest text not containing "}]") "}]"
 
-A :token:`TokCodeFragment` is nothing more than a multi-line string literal
-delimited by ``[{`` and ``}]``. It can break across lines.
+A :token:`TokCode` is nothing more than a multi-line string literal
+delimited by ``[{`` and ``}]``. It can break across lines and the
+line breaks are retained in the string.
 
 The current implementation accepts the following escape sequences::
 
@@ -254,7 +255,7 @@ high-level types (e.g., ``dag``). This flexibility allows you to describe a
 wide range of records conveniently and compactly.
 
 .. productionlist::
-   Type: "bit" | "int" | "string" | "code" | "dag"
+   Type: "bit" | "int" | "string" | "dag"
        :| "bits" "<" `TokInteger` ">"
        :| "list" "<" `Type` ">"
        :| `ClassID`
@@ -271,11 +272,6 @@ wide range of records conveniently and compactly.
     The ``string`` type represents an ordered sequence of characters of arbitrary
     length.
 
-``code``
-    The ``code`` type represents a code fragment. The values are the same as
-    those for the ``string`` type; the ``code`` type is provided just to indicate
-    the programmer's intention.
-
 ``bits<``\ *n*\ ``>``
     The ``bits`` type is a fixed-sized integer of arbitrary length *n* that
     is treated as separate bits. These bits can be accessed individually.
@@ -348,12 +344,12 @@ Simple values
 The :token:`SimpleValue` has a number of forms.
 
 .. productionlist::
-   SimpleValue: `TokInteger` | `TokString`+ | `TokCodeFragment`
+   SimpleValue: `TokInteger` | `TokString`+ | `TokCode`
 
-A value can be an integer literal, a string literal, or a code fragment
-literal. Multiple adjacent string literals are concatenated as in C/C++; the
-simple value is the concatenation of the strings. Code fragments become
-strings and then are indistinguishable from them.
+A value can be an integer literal, a string literal, or a code literal.
+Multiple adjacent string literals are concatenated as in C/C++; the simple
+value is the concatenation of the strings. Code literals become strings and
+are then indistinguishable from them.
 
 .. productionlist::
    SimpleValue2: "true" | "false"
@@ -616,14 +612,15 @@ name of a multiclass.
 
 .. productionlist::
    Body: ";" | "{" `BodyItem`* "}"
-   BodyItem: `Type` `TokIdentifier` ["=" `Value`] ";"
+   BodyItem: (`Type` | "code") `TokIdentifier` ["=" `Value`] ";"
            :| "let" `TokIdentifier` ["{" `RangeList` "}"] "=" `Value` ";"
            :| "defvar" `TokIdentifier` "=" `Value` ";"
 
 A field definition in the body specifies a field to be included in the class
 or record. If no initial value is specified, then the field's value is
 uninitialized. The type must be specified; TableGen will not infer it from
-the value.
+the value. The keyword ``code`` may be used to emphasize that the field
+has a string value that is code.
 
 The ``let`` form is used to reset a field to a new value. This can be done
 for fields defined directly in the body or fields inherited from

llvm/include/llvm/TableGen/Error.h

@@ -22,6 +22,7 @@ namespace llvm {
 void PrintNote(const Twine &Msg);
 void PrintNote(ArrayRef<SMLoc> NoteLoc, const Twine &Msg);
 
+LLVM_ATTRIBUTE_NORETURN void PrintFatalNote(const Twine &Msg);
 LLVM_ATTRIBUTE_NORETURN void PrintFatalNote(ArrayRef<SMLoc> ErrorLoc,
                                             const Twine &Msg);
 LLVM_ATTRIBUTE_NORETURN void PrintFatalNote(const Record *Rec,
@@ -37,6 +38,7 @@ void PrintError(const Twine &Msg);
 void PrintError(ArrayRef<SMLoc> ErrorLoc, const Twine &Msg);
 void PrintError(const char *Loc, const Twine &Msg);
 void PrintError(const Record *Rec, const Twine &Msg);
+void PrintError(const RecordVal *RecVal, const Twine &Msg);
 
 LLVM_ATTRIBUTE_NORETURN void PrintFatalError(const Twine &Msg);
 LLVM_ATTRIBUTE_NORETURN void PrintFatalError(ArrayRef<SMLoc> ErrorLoc,

llvm/include/llvm/TableGen/Record.h

@@ -58,7 +58,6 @@ class RecTy {
   enum RecTyKind {
     BitRecTyKind,
     BitsRecTyKind,
-    CodeRecTyKind,
     IntRecTyKind,
     StringRecTyKind,
     ListRecTyKind,
@@ -138,24 +137,6 @@ class BitsRecTy : public RecTy {
   bool typeIsA(const RecTy *RHS) const override;
 };
 
-/// 'code' - Represent a code fragment
-class CodeRecTy : public RecTy {
-  static CodeRecTy Shared;
-
-  CodeRecTy() : RecTy(CodeRecTyKind) {}
-
-public:
-  static bool classof(const RecTy *RT) {
-    return RT->getRecTyKind() == CodeRecTyKind;
-  }
-
-  static CodeRecTy *get() { return &Shared; }
-
-  std::string getAsString() const override { return "code"; }
-
-  bool typeIsConvertibleTo(const RecTy *RHS) const override;
-};
-
 /// 'int' - Represent an integer value of no particular size
 class IntRecTy : public RecTy {
   static IntRecTy Shared;
@@ -306,7 +287,6 @@ class Init {
     IK_FirstTypedInit,
     IK_BitInit,
     IK_BitsInit,
-    IK_CodeInit,
     IK_DagInit,
     IK_DefInit,
     IK_FieldInit,
@@ -597,16 +577,18 @@ class IntInit : public TypedInit {
 
 /// "foo" - Represent an initialization by a string value.
 class StringInit : public TypedInit {
-////  enum StringFormat {
-////    SF_String,  // Format as "text"
-////    SF_Code,    // Format as [{text}]
-////  };
+public:
+  enum StringFormat {
+    SF_String, // Format as "text"
+    SF_Code,   // Format as [{text}]
+  };
 
+private:
   StringRef Value;
-////  StringFormat Format;
+  StringFormat Format;
 
-  explicit StringInit(StringRef V)
-      : TypedInit(IK_StringInit, StringRecTy::get()), Value(V) {}
+  explicit StringInit(StringRef V, StringFormat Fmt)
+      : TypedInit(IK_StringInit, StringRecTy::get()), Value(V), Format(Fmt) {}
 
 public:
   StringInit(const StringInit &) = delete;
@@ -616,48 +598,25 @@ class StringInit : public TypedInit {
     return I->getKind() == IK_StringInit;
   }
 
-  static StringInit *get(StringRef);
-
-  StringRef getValue() const { return Value; }
-
-  Init *convertInitializerTo(RecTy *Ty) const override;
+  static StringInit *get(StringRef, StringFormat Fmt = SF_String);
 
-  bool isConcrete() const override { return true; }
-  std::string getAsString() const override { return "\"" + Value.str() + "\""; }
-
-  std::string getAsUnquotedString() const override {
-    return std::string(Value);
-  }
-
-  Init *getBit(unsigned Bit) const override {
-    llvm_unreachable("Illegal bit reference off string");
-  }
-};
-
-class CodeInit : public TypedInit {
-  StringRef Value;
-
-  explicit CodeInit(StringRef V)
-      : TypedInit(IK_CodeInit, static_cast<RecTy *>(CodeRecTy::get())),
-        Value(V) {}
-
-public:
-  CodeInit(const StringInit &) = delete;
-  CodeInit &operator=(const StringInit &) = delete;
-
-  static bool classof(const Init *I) {
-    return I->getKind() == IK_CodeInit;
+  static StringFormat determineFormat(StringFormat Fmt1, StringFormat Fmt2) {
+    return (Fmt1 == SF_Code || Fmt2 == SF_Code) ? SF_Code : SF_String;
   }
 
-  static CodeInit *get(StringRef);
-
   StringRef getValue() const { return Value; }
+  StringFormat getFormat() const { return Format; }  
+  bool hasCodeFormat() const { return Format == SF_Code; }
 
   Init *convertInitializerTo(RecTy *Ty) const override;
 
   bool isConcrete() const override { return true; }
+
   std::string getAsString() const override {
-    return "[{" + Value.str() + "}]";
+    if (Format == SF_String)
+      return "\"" + Value.str() + "\"";
+    else
+      return "[{" + Value.str() + "}]";
   }
 
   std::string getAsUnquotedString() const override {
@@ -1438,6 +1397,9 @@ class RecordVal {
   /// Get the type of the field value as a RecTy.
   RecTy *getType() const { return TyAndPrefix.getPointer(); }
 
+  /// Get the type of the field for printing purposes.
+  std::string getPrintType() const;
+
   /// Get the value of the field as an Init.
   Init *getValue() const { return Value; }
 
@@ -1675,11 +1637,6 @@ class Record {
   /// not a string and llvm::Optional() if the field does not exist.
   llvm::Optional<StringRef> getValueAsOptionalString(StringRef FieldName) const;
 
-  /// This method looks up the specified field and returns
-  /// its value as a string, throwing an exception if the field if the value is
-  /// not a code block and llvm::Optional() if the field does not exist.
-  llvm::Optional<StringRef> getValueAsOptionalCode(StringRef FieldName) const;
-
   /// This method looks up the specified field and returns
   /// its value as a BitsInit, throwing an exception if the field does not exist
   /// or if the value is not the right type.

llvm/include/llvm/TableGen/SearchableTable.td

@@ -67,9 +67,13 @@ class GenericTable {
   // List of the names of fields of collected records that contain the data for
   // table entries, in the order that is used for initialization in C++.
   //
-  // For each field of the table named XXX, TableGen will look for a value
-  // called TypeOf_XXX and use that as a more detailed description of the
-  // type of the field if present. This is required for fields whose type
+  // TableGen needs to know the type of the fields so that it can format
+  // the initializers correctly. It can infer the type of bit, bits, string,
+  // Intrinsic, and Instruction values. 
+  //
+  // For each field of the table named xxx, TableGen will look for a field
+  // named TypeOf_xxx and use that as a more detailed description of the
+  // type of the field. This is required for fields whose type
   // cannot be deduced automatically, such as enum fields. For example:
   //
   //   def MyEnum : GenericEnum {
@@ -85,15 +89,15 @@ class GenericTable {
   //   def MyTable : GenericTable {
   //     let FilterClass = "MyTableEntry";
   //     let Fields = ["V", ...];
-  //     GenericEnum TypeOf_V = MyEnum;
+  //     string TypeOf_V = "MyEnum";
   //   }
   //
-  // Fields of type bit, bits<N>, string, Intrinsic, and Instruction (or
-  // derived classes of those) are supported natively.
+  // If a string field was initialized with a code literal, TableGen will
+  // emit the code verbatim. However, if a string field was initialized
+  // in some other way, but should be interpreted as code, then a TypeOf_xxx
+  // field is necessary, with a value of "code":
   //
-  // Additionally, fields of type `code` can appear, where the value is used
-  // verbatim as an initializer. However, these fields cannot be used as
-  // search keys.
+  //     string TypeOf_Predicate = "code";
   list<string> Fields;
 
   // (Optional) List of fields that make up the primary key.