path: root/testsuite/awk.tests

#!/bin/sh

# Copyright 2007 by Denys Vlasenko <vda.linux@googlemail.com>
# Licensed under GPLv2, see file LICENSE in this source tree.

. ./testing.sh

# testing "description" "command" "result" "infile" "stdin"

testing "awk -F case 0" "awk -F '[#]' '{ print NF }'" ""    "" ""
testing "awk -F case 1" "awk -F '[#]' '{ print NF }'" "0\n" "" "\n"
testing "awk -F case 2" "awk -F '[#]' '{ print NF }'" "2\n" "" "#\n"
testing "awk -F case 3" "awk -F '[#]' '{ print NF }'" "3\n" "" "#abc#\n"
testing "awk -F case 4" "awk -F '[#]' '{ print NF }'" "3\n" "" "#abc#zz\n"
testing "awk -F case 5" "awk -F '[#]' '{ print NF }'" "4\n" "" "#abc##zz\n"
testing "awk -F case 6" "awk -F '[#]' '{ print NF }'" "4\n" "" "z#abc##zz\n"
testing "awk -F case 7" "awk -F '[#]' '{ print NF }'" "5\n" "" "z##abc##zz\n"

# conditions and operators
testing "awk if operator == "  "awk 'BEGIN{if(23==23) print \"foo\"}'" "foo\n" "" ""
testing "awk if operator != "  "awk 'BEGIN{if(23!=23) print \"bar\"}'" ""      "" ""
testing "awk if operator >= "  "awk 'BEGIN{if(23>=23) print \"foo\"}'" "foo\n" "" ""
testing "awk if operator < "   "awk 'BEGIN{if(2 < 13) print \"foo\"}'" "foo\n" "" ""
testing "awk if string == "    "awk 'BEGIN{if(\"a\"==\"ab\") print \"bar\"}'" "" "" ""

# 4294967295 = 0xffffffff
testing "awk bitwise op"  "awk '{ print or(4294967295,1) }'" "4294967295\n" "" "\n"

# we were testing for a non-empty body when deciding if a function was
# defined or not. The testcase below caused:
# awk: cmd. line:8: Call to undefined function
prg='
function empty_fun(count) {
  # empty
}
END {
  i=1
  print "L" i "\n"
  empty_fun(i + i + ++i)
  print "L" i "\n"
}'
testing "awk handles empty function f(arg){}" \
	"awk '$prg'" \
	"L1\n\nL2\n\n" \
	"" ""

prg='
function outer_fun() {
  return 1
}
END {
  i=1
  print "L" i "\n"
  i += outer_fun()
  print "L" i "\n"
}'
testing "awk properly handles function from other scope" \
	"awk '$prg'" \
	"L1\n\nL2\n\n" \
	"" ""

prg='
END {
  i=1
  print "L" i "\n"
  i + trigger_error_fun()
  print "L" i "\n"
}'
testing "awk properly handles undefined function" \
	"awk '$prg' 2>&1" \
	"L1\n\nawk: cmd. line:5: Call to undefined function\n" \
	"" ""


optional DESKTOP
testing "awk hex const 1" "awk '{ print or(0xffffffff,1) }'" "4294967295\n" "" "\n"
testing "awk hex const 2" "awk '{ print or(0x80000000,1) }'" "2147483649\n" "" "\n"
testing "awk oct const"   "awk '{ print or(01234,1) }'"      "669\n"         "" "\n"
SKIP=

# check that "hex/oct integer" heuristic doesn't kick in on 00NN.NNN
testing "awk floating const with leading zeroes" \
	"awk '{ printf \"%f %f\n\", \"000.123\", \"009.123\" }'" \
	"0.123000 9.123000\n" \
	"" "\n"

# long field seps requiring regex
testing "awk long field sep" \
	"awk -F-- '{ print NF, length(\$NF), \$NF }'" \
	"2 0 \n3 0 \n4 0 \n5 0 \n" \
	"" \
	"a--\na--b--\na--b--c--\na--b--c--d--"

testing "awk -F handles escapes" "awk -F'\\x21' '{print \$1}'" \
	"a\n" \
	"" \
	"a!b\n"

# '@(samp|code|file)\{' is an invalid extended regex (unmatched '{'),
# but gawk 3.1.5 does not bail out on it.
testing "awk gsub falls back to non-extended-regex" \
	"awk 'gsub(\"@(samp|code|file)\{\",\"\");'; echo \$?" "0\n" "" "Hi\n"

optional TAR BUNZIP2 FEATURE_SEAMLESS_BZ2
test x"$SKIP" != x"1" && tar xjf awk_t1.tar.bz2
testing "awk 'gcc build bug'" \
	"awk -f awk_t1_opt-functions.awk -f awk_t1_opth-gen.awk <awk_t1_input | md5sum" \
	"f842e256461a5ab1ec60b58d16f1114f  -\n" \
	"" ""
rm -rf awk_t1_* 2>/dev/null
SKIP=

Q='":"'

testing "awk NF in BEGIN" \
	"awk 'BEGIN { print ${Q} NF ${Q} \$0 ${Q} \$1 ${Q} \$2 ${Q} }'" \
	":0::::\n" \
	"" ""

prg='
function b(tmp) {
	tmp = 0;
	print "" tmp; #this line causes the bug
	return tmp;
}
function c(tmpc) {
	tmpc = b(); return tmpc;
}
BEGIN {
	print (c() ? "string" : "number");
}'
testing "awk string cast (bug 725)" \
	"awk '$prg'" \
	"0\nnumber\n" \
	"" ""

testing "awk handles whitespace before array subscript" \
	"awk 'BEGIN { arr [3] = 1; print arr [3] }'" "1\n" "" ""

# GNU awk 3.1.5's "print ERRNO" prints "No such file or directory" instead of "2",
# do we need to emulate that as well?
testing "awk handles non-existing file correctly" \
	"awk 'BEGIN { getline line <\"doesnt_exist\"; print ERRNO; ERRNO=0; close(\"doesnt_exist\"); print ERRNO; print \"Ok\" }'" \
	"2\n0\nOk\n" "" ""

prg='
BEGIN {
  u["a"]=1
  u["b"]=1
  u["c"]=1
  v["d"]=1
  v["e"]=1
  v["f"]=1
  for (l in u) {
    print "outer1", l;
    for (l in v) {
      print " inner", l;
    }
    print "outer2", l;
  }
  print "end", l;
  l="a"
  exit;
}'
testing "awk nested loops with the same variable" \
	"awk '$prg'" \
	"\
outer1 a
 inner d
 inner e
 inner f
outer2 f
outer1 b
 inner d
 inner e
 inner f
outer2 f
outer1 c
 inner d
 inner e
 inner f
outer2 f
end f
" \
	"" ""

prg='
BEGIN {
  u["a"]=1
  u["b"]=1
  u["c"]=1
  v["d"]=1
  v["e"]=1
  v["f"]=1
  for (l in u) {
    print "outer1", l;
    for (l in v) {
      print " inner", l;
      break;
    }
    print "outer2", l;
  }
  print "end", l;
  l="a"
  exit;
}'
# It's not just buggy, it enters infinite loop. Thus disabled
false && test x"$SKIP_KNOWN_BUGS" = x"" && testing "awk nested loops with the same variable and break" \
	"awk '$prg'" \
	"\
outer1 a
 inner d
outer2 d
outer1 b
 inner d
outer2 d
outer1 c
 inner d
outer2 d
end d
" \
	"" ""

prg='
function f() {
  for (l in v) {
    print " inner", l;
    return;
  }
}

BEGIN {
  u["a"]=1
  u["b"]=1
  u["c"]=1
  v["d"]=1
  v["e"]=1
  v["f"]=1
  for (l in u) {
    print "outer1", l;
    f();
    print "outer2", l;
  }
  print "end", l;
  l="a"
  exit;
}'
# It's not just buggy, it enters infinite loop. Thus disabled
false && test x"$SKIP_KNOWN_BUGS" = x"" && testing "awk nested loops with the same variable and return" \
	"awk '$prg'" \
	"\
outer1 a
 inner d
outer2 d
outer1 b
 inner d
outer2 d
outer1 c
 inner d
outer2 d
end d
" \
	"" ""

prg='
BEGIN{
cnt = 0
a[cnt] = "zeroth"
a[++cnt] = "first"
delete a[cnt--]
print cnt
print "[0]:" a[0]
print "[1]:" a[1]
}'
testing "awk 'delete a[v--]' evaluates v-- once" \
	"awk '$prg'" \
	"\
0
[0]:zeroth
[1]:
" \
	"" ""

testing "awk func arg parsing 1" \
	"awk 'func f(,) { }' 2>&1" "awk: cmd. line:1: Unexpected token\n" "" ""

testing "awk func arg parsing 2" \
	"awk 'func f(a,,b) { }' 2>&1" "awk: cmd. line:1: Unexpected token\n" "" ""

testing "awk func arg parsing 3" \
	"awk 'func f(a,) { }' 2>&1" "awk: cmd. line:1: Unexpected token\n" "" ""

testing "awk func arg parsing 4" \
	"awk 'func f(a b) { }' 2>&1" "awk: cmd. line:1: Unexpected token\n" "" ""

testing "awk handles empty ()" \
	"awk 'BEGIN {print()}' 2>&1" "awk: cmd. line:1: Empty sequence\n" "" ""

testing "awk FS assignment" "awk '{FS=\":\"; print \$1}'" \
	"a:b\ne\n" \
	"" \
	"a:b c:d\ne:f g:h"

optional FEATURE_AWK_LIBM
testing "awk large integer" \
	"awk 'BEGIN{n=(2^31)-1; print n, int(n), n%1, ++n, int(n), n%1}'" \
	"2147483647 2147483647 0 2147483648 2147483648 0\n" \
	"" ""
SKIP=

testing "awk length(array)" \
	"awk 'BEGIN{ A[1]=2; A[\"qwe\"]=\"asd\"; print length(A)}'" \
	"2\n" \
	"" ""

testing "awk length()" \
	"awk '{print length; print length(); print length(\"qwe\"); print length(99+9)}'" \
	"3\n3\n3\n3\n" \
	"" "qwe"

testing "awk print length, 1" \
	"awk '{ print length, 1 }'" \
	"0 1\n" \
	"" "\n"

testing "awk print length 1" \
	"awk '{ print length 1 }'" \
	"01\n" \
	"" "\n"

testing "awk length == 0" \
	"awk 'length == 0 { print \"foo\" }'" \
	"foo\n" \
	"" "\n"

testing "awk if (length == 0)" \
	"awk '{ if (length == 0) { print \"bar\" } }'" \
	"bar\n" \
	"" "\n"

testing "awk -f and ARGC" \
	"awk -f - input" \
	"re\n2\n" \
	"do re mi\n" \
	'{print $2; print ARGC;}' \

optional FEATURE_AWK_GNU_EXTENSIONS
testing "awk -e and ARGC" \
	"awk -e '{print \$2; print ARGC;}' input" \
	"re\n2\n" \
	"do re mi\n" \
	""
SKIP=

# The examples are in fact not valid awk programs (break/continue
# can only be used inside loops).
# But we do accept them outside of loops.
# We had a bug with misparsing "break ; else" sequence.
# Test that *that* bug is fixed, using simplest possible scripts:
testing "awk break" \
	"awk -f - 2>&1; echo \$?" \
	"0\n" \
	"" \
	'BEGIN { if (1) break; else a = 1 }'
testing "awk continue" \
	"awk -f - 2>&1; echo \$?" \
	"0\n" \
	"" \
	'BEGIN { if (1) continue; else a = 1 }'

optional FEATURE_AWK_GNU_EXTENSIONS
testing "awk handles invalid for loop" \
	"awk -e '{ for() }' 2>&1" "awk: cmd. line:1: Unexpected token\n" "" ""
SKIP=

optional FEATURE_AWK_GNU_EXTENSIONS
testing "awk handles colon not preceded by ternary" \
	"awk -e foo:bar: 2>&1" "awk: cmd. line:1: Unexpected token\n" "" ""
SKIP=

optional FEATURE_AWK_GNU_EXTENSIONS
testing "awk errors on missing delete arg" \
	"awk -e '{delete}' 2>&1" "awk: cmd. line:1: Too few arguments\n" "" ""
SKIP=

# testing "description" "command" "result" "infile" "stdin"
testing 'awk negative field access' \
	'awk 2>&1 -- '\''{ $(-1) }'\' \
	"awk: cmd. line:1: Access to negative field\n" \
	'' \
	'anything'

# was misinterpreted as (("str"++) i) instead of ("str" (++i))
# (and was executed: "str"++ is "0", thus concatenating "0" and "1"):
testing 'awk do not allow "str"++' \
	'awk -v i=1 "BEGIN {print \"str\" ++i}"' \
	"str2\n" \
	'' \
	'anything'

# gawk compat: FS regex matches only non-empty separators:
# with -*, the splitting is NOT f o o b a r, but foo bar:
testing 'awk FS regex which can match empty string' \
	"awk -F '-*' '{print \$1 \"-\" \$2 \"=\" \$3 \"*\" \$4}'" \
	"foo-bar=*\n" \
	'' \
	'foo--bar'

# last+1 field should be empty (had a bug where it wasn't)
testing 'awk $NF is empty' \
	"awk -F '=+' '{print \$NF}'" \
	"\n" \
	'' \
	'a=====123='

exit $FAILCOUNT