Disallow var syntax in string interpolation (#32948)

The var"##" syntax should be disabled in string interpolation.

Disallow `var` syntax in command interpolations

This is special cased for compatibility. A more general fix would be to
make cmd interpolation syntax exactly the same as string interpolation.

(cherry picked from commit 050160c)

Author: c42f

base/shell.jl

@@ -70,7 +70,14 @@ function shell_parse(str::AbstractString, interpolate::Bool=true;
             isempty(st) && error("\$ right before end of command")
             stpos, c = popfirst!(st)
             isspace(c) && error("space not allowed right after \$")
-            ex, j = Meta.parse(s,stpos,greedy=false)
+            if startswith(SubString(s, stpos), "var\"")
+                # Disallow var"#" syntax in cmd interpolations.
+                # TODO: Allow only identifiers after the $ for consistency with
+                # string interpolation syntax (see #3150)
+                ex, j = :var, stpos+3
+            else
+                ex, j = Meta.parse(s,stpos,greedy=false)
+            end
             last_parse = (stpos:prevind(s, j)) .+ s.offset
             update_arg(ex);
             s = SubString(s, j)

src/julia-parser.scm

@@ -2134,11 +2134,11 @@
   (let* ((p (ts:port s))
          (c (peek-char p)))
     (cond ((identifier-start-char? c)
-           (let* ((atom (parse-atom s))
+           (let* ((t (require-token s))
                   (c (peek-char p)))
              (if (ends-interpolated-atom? c)
-                 atom
-                 (error (string "interpolated variable $" atom " ends with invalid character \"" c "\"; use \"$(" atom ")\" instead.")))))
+                 (take-token s)
+                 (error (string "interpolated variable $" t " ends with invalid character \"" c "\"; use \"$(" t ")\" instead.")))))
           ((eqv? c #\()
            (read-char p)
            (let ((ex (parse-eq* s))
@@ -2285,14 +2285,20 @@
                       (if (closing-token? t)
                           (error (string "unexpected \"" (take-token s) "\"")))))
            (take-token s)
-           (if (and (eq? t 'var) (eqv? (peek-token s) #\") (not (ts:space? s)))
+           (if (and (eq? t 'var)
+                    (if (or (ts:pbtok s) (ts:last-tok s))
+                        (and (eqv? (peek-token s) #\") (not (ts:space? s)))
+                        ;; Hack: avoid peek-token if possible to preserve
+                        ;; (io.pos (ts:port s)) for non-greedy Meta.parse
+                        (eqv? (peek-char (ts:port s)) #\")))
              (begin
                ;; var"funky identifier" syntax
-               (take-token s)
+               (peek-token s)
+               (take-token s) ;; leading "
                (let ((str (parse-raw-literal s #\"))
                      (nxt (peek-token s)))
                  (if (and (symbol? nxt) (not (operator? nxt)) (not (ts:space? s)))
-                   (error (string "suffix not allowed after `var\"" str "\"`")))
+                     (error (string "suffix not allowed after `var\"" str "\"`")))
                  (symbol str)))
              t))
 

test/spawn.jl

@@ -461,6 +461,15 @@ let c = setenv(`x`, "A"=>true)
     @test_throws ArgumentError `"$c "`
 end
 
+# Interaction of cmd parsing with var syntax (#32408)
+let var = "x", vars="z"
+    @test `ls $var` == Cmd(["ls", "x"])
+    @test `ls $vars` == Cmd(["ls", "z"])
+    @test `ls $var"y"` == Cmd(["ls", "xy"])
+    @test `ls "'$var'"` == Cmd(["ls", "'x'"])
+    @test `ls $var "y"` == Cmd(["ls", "x", "y"])
+end
+
 # equality tests for AndCmds
 @test Base.AndCmds(`$echocmd abc`, `$echocmd def`) == Base.AndCmds(`$echocmd abc`, `$echocmd def`)
 @test Base.AndCmds(`$echocmd abc`, `$echocmd def`) != Base.AndCmds(`$echocmd abc`, `$echocmd xyz`)

test/syntax.jl

@@ -1922,3 +1922,9 @@ end
 @test Base.remove_linenums!(Meta.parse("try a catch var\"#\" b end")) ==
       Expr(:try, Expr(:block, :a), Symbol("#"), Expr(:block, :b))
 @test Meta.parse("(var\"function\" = 1,)") == Expr(:tuple, Expr(:(=), Symbol("function"), 1))
+# Non-standard identifiers require parens for string interpolation
+@test Meta.parse("\"\$var\\\"#\\\"\"") == Expr(:string, :var, "\"#\"")
+@test Meta.parse("\"\$(var\"#\")\"") == Expr(:string, Symbol("#"))
+# Stream positioning after parsing var
+@test Meta.parse("var'", 1, greedy=false) == (:var, 4)
+