libnixf: suggest adding builtins. for known primops (#720)

Author: inclyc

flake.nix

@@ -35,7 +35,7 @@
             ;
           nixComponents = nixVersions.nixComponents_2_30;
           llvmPackages = llvmPackages_19;
-          nixf = callPackage ./libnixf { };
+          nixf = callPackage ./libnixf { inherit (nixComponents) nix-expr; };
           nixt = callPackage ./libnixt { inherit nixComponents; };
           nixd = callPackage ./nixd {
             inherit nixComponents nixf nixt;

libnixf/default.nix

@@ -9,6 +9,7 @@
   boost,
   nlohmann_json,
   python312,
+  nix-expr,
 }:
 
 stdenv.mkDerivation {
@@ -42,6 +43,7 @@ stdenv.mkDerivation {
     gtest
     boost
     nlohmann_json
+    nix-expr
   ];
 
   meta = {

libnixf/include/nixf/Sema/PrimOpInfo.h

@@ -0,0 +1,52 @@
+#include <cstdint>
+#include <map>
+#include <string>
+#include <vector>
+
+namespace nixf {
+
+/**
+ * Info about a primitive operation, but omit the implementation.
+ */
+struct PrimOpInfo {
+  /**
+   * Names of the parameters of a primop, for primops that take a
+   * fixed number of arguments to be substituted for these parameters.
+   */
+  std::vector<std::string> Args;
+
+  /**
+   * Aritiy of the primop.
+   *
+   * If `args` is not empty, this field will be computed from that
+   * field instead, so it doesn't need to be manually set.
+   */
+  size_t Arity = 0;
+
+  /**
+   * Optional free-form documentation about the primop.
+   */
+  std::string Doc;
+
+  /**
+   * If true, this primop is not exposed to the user.
+   */
+  bool Internal;
+};
+
+extern std::map<std::string, nixf::PrimOpInfo> PrimOpsInfo;
+
+/// \brief Result of looking up a primop by name.
+enum class PrimopLookupResult : std::uint8_t {
+  /// The primop was found with an exact match.
+  Found,
+  /// The primop was found, but needs "builtin." prefix.
+  PrefixedFound,
+  /// The primop was not found.
+  NotFound,
+};
+
+/// \brief Look up information about a global primop by name.
+PrimopLookupResult lookupGlobalPrimOpInfo(const std::string &Name);
+
+} // namespace nixf

libnixf/include/nixf/Sema/VariableLookup.h

@@ -115,6 +115,7 @@ class VariableLookupAnalysis {
     Undefined,
     FromWith,
     Defined,
+    PrimOp,
     NoSuchVar,
   };
 

libnixf/meson.build

@@ -4,6 +4,7 @@ project( 'nixf'
        , version: 'nightly'
        )
 
+nix_expr = dependency('nix-expr')
 nlohmann_json = dependency('nlohmann_json')
 boost = dependency('boost')
 gtest = dependency('gtest')

libnixf/src/Basic/PrimOpsInfoGen.cpp

@@ -0,0 +1,43 @@
+#include <nix/expr/eval.hh>
+#include <nix/expr/primops.hh>
+
+#include <iostream>
+
+int main(int argc, char **argv) {
+  std::freopen(argv[1], "w", stdout);
+  nix::initGC();
+  // Generate .cpp file that inject all primops into a custom vector.
+  // This is used by the language server to provide documentation,
+  // but omit the implementation of primops.
+  std::cout << "#include <nixf/Sema/PrimOpInfo.h>" << "\n\n";
+  std::cout << "std::map<std::string, nixf::PrimOpInfo> nixf::PrimOpsInfo = {"
+            << '\n';
+  for (const auto &PrimOp : nix::RegisterPrimOp::primOps()) {
+    std::cout << "  {" << "\"" << PrimOp.name << "\", {\n";
+
+    // .args
+    std::cout << "    .Args = {";
+    for (size_t I = 0; I < PrimOp.args.size(); ++I) {
+      std::cout << "\"" << PrimOp.args[I] << "\"";
+      if (I + 1 < PrimOp.args.size()) {
+        std::cout << ", ";
+      }
+    }
+    std::cout << "},\n";
+
+    // .arity
+    std::cout << "    .Arity = " << PrimOp.arity << ",\n";
+
+    // .doc
+    std::cout << "    .Doc = R\"xabc(" << (PrimOp.doc ? PrimOp.doc : "")
+              << ")xabc\",\n";
+
+    // .internal
+    std::cout << "    .Internal = " << (PrimOp.internal ? "true" : "false")
+              << ",\n";
+
+    std::cout << "  }},\n";
+  }
+  std::cout << "};\n";
+  return 0;
+}

libnixf/src/Basic/diagnostic.py

@@ -231,4 +231,13 @@ class Diagnostic(TypedDict):
         "severity": "Warning",
         "message": "unused `with` expression",
     },
+    # Primary Operation (PrimOp) related diagnostics
+    # We call this "builtin" in messages to match Nix's terminology
+    # "PrimOp" is specific to Nix/nixf implementation
+    {
+        "sname": "sema-primop-needs-prefix",
+        "cname": "PrimOpNeedsPrefix",
+        "severity": "Error",
+        "message": "this is not a prelude builtin, the `builtins.` prefix is required",
+    },
 ]

libnixf/src/Sema/PrimOpLookup.cpp

@@ -0,0 +1,17 @@
+#include "nixf/Sema/PrimOpInfo.h"
+
+using namespace nixf;
+
+PrimopLookupResult nixf::lookupGlobalPrimOpInfo(const std::string &Name) {
+  if (PrimOpsInfo.contains(Name)) {
+    return PrimopLookupResult::Found;
+  }
+
+  // Prefix the name with "__", and check again.
+  std::string PrefixedName = "__" + Name;
+  if (PrimOpsInfo.contains(PrefixedName)) {
+    return PrimopLookupResult::PrefixedFound;
+  }
+
+  return PrimopLookupResult::NotFound;
+}

libnixf/src/Sema/VariableLookup.cpp

@@ -2,6 +2,7 @@
 #include "nixf/Basic/Diagnostic.h"
 #include "nixf/Basic/Nodes/Attrs.h"
 #include "nixf/Basic/Nodes/Lambda.h"
+#include "nixf/Sema/PrimOpInfo.h"
 
 using namespace nixf;
 
@@ -108,11 +109,28 @@ void VariableLookupAnalysis::lookupVar(const ExprVar &Var,
     }
     Results.insert({&Var, LookupResult{LookupResultKind::FromWith, Def}});
   } else {
-    // Otherwise, this variable is undefined.
-    Results.insert({&Var, LookupResult{LookupResultKind::Undefined, nullptr}});
-    Diagnostic &Diag =
-        Diags.emplace_back(Diagnostic::DK_UndefinedVariable, Var.range());
-    Diag << Var.id().name();
+    // Check if this is a primop.
+    switch (lookupGlobalPrimOpInfo(Name)) {
+    case PrimopLookupResult::Found:
+      Results.insert({&Var, LookupResult{LookupResultKind::PrimOp, nullptr}});
+      break;
+    case PrimopLookupResult::PrefixedFound: {
+      Diagnostic &D =
+          Diags.emplace_back(Diagnostic::DK_PrimOpNeedsPrefix, Var.range());
+      D.fix("use `builtins.` prefix")
+          .edit(TextEdit::mkInsertion(Var.range().lCur(), "builtins."));
+      Results.insert({&Var, LookupResult{LookupResultKind::PrimOp, nullptr}});
+      break;
+    }
+    case PrimopLookupResult::NotFound:
+      // Otherwise, this variable is undefined.
+      Results.insert(
+          {&Var, LookupResult{LookupResultKind::Undefined, nullptr}});
+      Diagnostic &Diag =
+          Diags.emplace_back(Diagnostic::DK_UndefinedVariable, Var.range());
+      Diag << Var.id().name();
+      break;
+    }
   }
 }
 
@@ -344,129 +362,20 @@ void VariableLookupAnalysis::dfs(const Node &Root,
 void VariableLookupAnalysis::runOnAST(const Node &Root) {
   // Create a basic env
   DefBuilder DB;
-  std::vector<std::string> Builtins{
-      "__add",
-      "__fetchurl",
-      "__isFloat",
-      "__seq",
-      "break",
-      "__addDrvOutputDependencies",
-      "__filter",
-      "__isFunction",
-      "__sort",
+  std::vector<std::string> BaseEnv{
       "builtins",
-      "__addErrorContext",
-      "__filterSource",
-      "__isInt",
-      "__split",
-      "derivation",
-      "__all",
-      "__findFile",
-      "__isList",
-      "__splitVersion",
-      "derivationStrict",
-      "__any",
-      "__flakeRefToString",
-      "__isPath",
-      "__storeDir",
-      "dirOf",
-      "__appendContext",
-      "__floor",
-      "__isString",
-      "__storePath",
-      "false",
-      "__attrNames",
-      "__foldl'",
-      "__langVersion",
-      "__stringLength",
-      "fetchGit",
-      "__attrValues",
-      "__fromJSON",
-      "__length",
-      "__sub",
-      "fetchMercurial",
-      "__bitAnd",
-      "__functionArgs",
-      "__lessThan",
-      "__substring",
-      "fetchTarball",
-      "__bitOr",
-      "__genList",
-      "__listToAttrs",
-      "__tail",
-      "fetchTree",
-      "__bitXor",
-      "__genericClosure",
-      "__mapAttrs",
-      "__toFile",
-      "fromTOML",
-      "__catAttrs",
-      "__getAttr",
-      "__match",
-      "__toJSON",
-      "import",
-      "__ceil",
-      "__getContext",
-      "__mul",
-      "__toPath",
-      "isNull",
-      "__compareVersions",
-      "__getEnv",
-      "__nixPath",
-      "__toXML",
-      "map",
-      "__concatLists",
-      "__getFlake",
-      "__nixVersion",
-      "__trace",
-      "null",
-      "__concatMap",
-      "__groupBy",
-      "__parseDrvName",
-      "__traceVerbose",
-      "placeholder",
-      "__concatStringsSep",
-      "__hasAttr",
-      "__parseFlakeRef",
-      "__tryEval",
-      "removeAttrs",
-      "__convertHash",
-      "__hasContext",
-      "__partition",
-      "__typeOf",
-      "scopedImport",
-      "__currentSystem",
-      "__hashFile",
-      "__path",
-      "__unsafeDiscardOutputDependency",
-      "throw",
-      "__currentTime",
-      "__hashString",
-      "__pathExists",
-      "__unsafeDiscardStringContext",
-      "toString",
-      "__deepSeq",
-      "__head",
-      "__readDir",
-      "__unsafeGetAttrPos",
-      "true",
-      "__div",
-      "__intersectAttrs",
-      "__readFile",
-      "__zipAttrsWith",
-      "__elem",
-      "__isAttrs",
-      "__readFileType",
-      "abort",
-      "__elemAt",
-      "__isBool",
-      "__replaceStrings",
-      "baseNameOf",
       // This is an undocumented keyword actually.
       "__curPos",
   };
 
-  for (const auto &Builtin : Builtins)
+  for (const auto &[Name, Info] : PrimOpsInfo) {
+    if (!Info.Internal && !Name.starts_with("__")) {
+      // Only add non-internal primops without "__" prefix.
+      BaseEnv.push_back(Name);
+    }
+  }
+
+  for (const auto &Builtin : BaseEnv)
     DB.addBuiltin(Builtin);
 
   auto Env = std::make_shared<EnvNode>(nullptr, DB.finish(), nullptr);

libnixf/src/meson.build

@@ -35,6 +35,21 @@ tokens_kinds_inc = custom_target(
     install_dir: 'include/nixf/Parse/'
 )
 
+
+prim_ops_info_gen = executable(
+    'PrimOpsInfo',
+    'Basic/PrimOpsInfoGen.cpp',
+    dependencies: [ nix_expr ],
+    install: false,
+)
+
+
+builtins_cpp = custom_target(
+    input: [ prim_ops_info_gen ],
+    output: 'PrimOpsInfo.cpp',
+    command: [ prim_ops_info_gen,  '@OUTPUT@' ],
+)
+
 libnixf = library(
     'nixf',
     'Basic/Nodes.cpp',
@@ -49,12 +64,14 @@ libnixf = library(
     'Parse/ParseStrings.cpp',
     'Parse/ParseSupport.cpp',
     'Sema/ParentMap.cpp',
+    'Sema/PrimOpLookup.cpp',
     'Sema/SemaActions.cpp',
     'Sema/VariableLookup.cpp',
     diagnostic_enum_h,
     diagnostic_cpp,
     tokens_h,
     tokens_kinds_inc,
+    builtins_cpp,
     include_directories: libnixf_inc,
     dependencies: libnixf_deps,
     install: true,