Pike Language Reference

Pike Language Reference

Overview

Pike is a dynamic, bytecode-compiled, object-oriented language with C-like syntax. Target version: 8.0.1116. It features strong typing with type inference, first-class functions/closures, and a rich standard library. Pike compiles to bytecode and runs on its own virtual machine.

Key Concepts

• Value types (int, float, string) are copied on assignment.
• Reference types (array, mapping, multiset, object, program, function) share underlying data on assignment.
• Strings use copy-on-write: assigning to a new variable creates an independent copy, but individual characters can be modified via index assignment (s[0] = 'H'). Use String.Buffer for efficient concatenation.
• Arrays, mappings, multisets, objects, functions, programs, and types use identity comparison with ==. Use equal() for structural/deep comparison.
• Integer division rounds toward negative infinity (floor division), not toward zero.
• switch cases DO fall through — use break to prevent (like C/Java).
• foreach iterates by value with optional index; iterate strings as character codes (int).

Rules

Rule: Literal Syntax

Pike uses curly-brace forms for collection literals, not square brackets or curly braces alone.

WRONG:

array a = [1, 2, 3];
mapping m = {"key": "value"};
multiset s = {"a", "b"};

CORRECT:

array a = ({1, 2, 3});
mapping m = (["key": "value"]);
multiset s = (< "a", "b" >);

Rule: String Mutation and Concatenation

Strings use copy-on-write: assigning a string to another variable creates an independent copy, so index assignment on one variable does not affect the other. However, String.Buffer should still be used for efficient concatenation in loops.

WRONG:

string result = "";
for (int i = 0; i < 1000; i++)
  result += "x"; // O(n^2): allocates new string each iteration

CORRECT:

String.Buffer buf = String.Buffer();
for (int i = 0; i < 1000; i++)
  buf->add("x");
string result = buf->get();

Rule: Array Identity vs Equality

The == operator on arrays checks object identity (same array object), not structural equality. Use equal() for deep comparison.

WRONG:

array a = ({1, 2, 3});
array b = ({1, 2, 3});
if (a == b) // false — different array objects
  write("same\n");

CORRECT:

array a = ({1, 2, 3});
array b = ({1, 2, 3});
if (equal(a, b)) // true — same contents
  write("same\n");

Rule: Reference Semantics for Pointer Types

Arrays, mappings, multisets, objects, and programs are reference types. Assigning to a variable or passing to a function shares the same underlying data. Modifying through one alias affects all others.

WRONG:

array original = ({1, 2, 3});
array copy = original;
copy[0] = 99;
// original is now ({99, 2, 3}) — NOT ({1, 2, 3})

CORRECT:

array original = ({1, 2, 3});
array copy = copy_value(original); // deep copy
copy[0] = 99;
// original remains ({1, 2, 3})

Rule: Integer Division Rounding

Integer division rounds toward negative infinity (floor division), not toward zero. This differs from C/Java.

WRONG:

// Assuming C-style truncation toward zero:
int a = -8 / 3;  // expecting -2 (wrong)

CORRECT:

int a = -8 / 3;  // -3 — rounds toward -inf
int b = 8 / 3;   // 2 — same as C for positive numbers
// If truncation toward zero is needed, cast:
int c = (int)((float)-8 / 3); // -2

Rule: Constructor and Destructor Naming

Constructors are named create(), not __init__ or __construct. Destructors are destroy().

WRONG:

class Foo {
  void __init() { }     // not a constructor
  void __construct() { } // not a constructor
}

CORRECT:

class Foo {
  void create(string name) {
    // constructor — called on Foo("bar")
  }
  void destroy() {
    // destructor — called when garbage collected
  }
}

Rule: Inheritance Syntax

Use inherit keyword. Call parent methods with :: prefix. Use this_program:: for disambiguation.

WRONG:

class Child extends Parent { }
class Child : Parent { }
void parent_method() { super(); }

CORRECT:

class Child {
  inherit Parent;
  void create() {
    ::create(); // call parent constructor
  }
  void method() {
    ::method(); // call parent method
  }
}

Rule: Switch Fall-Through

Pike switch cases DO fall through (like C/Java). Use break to prevent fall-through. Missing break is a common bug.

WRONG (missing break — will fall through to next case):

switch (x) {
  case 1:
    do_one();
  case 2:
    do_two();
  // BUG: falls through into do_range()!
  case 3..5:
    do_range();
}

CORRECT:

switch (x) {
  case 1:
    do_one();
    break;
  case 2:
    do_two();
    break;
  case 3..5:
    do_range();
    break;
  default:
    do_default();
    break;
}
// Use stacked case labels for shared handling:
switch (x) {
  case 1:
  case 2:
  case 3:
    handle_one_two_three();
    break;
}

Rule: Foreach Syntax

foreach iterates by value. Use the three-argument form for index access. Strings iterate as character codes (int).

WRONG:

foreach (arr as val) { }        // PHP syntax
for (val in arr) { }            // Python syntax
foreach (string s in "hello") { // strings don't iterate as strings

CORRECT:

foreach (arr; int idx; mixed val) { }
foreach (mapping; mixed key; mixed val) { }
foreach (multiset; mixed key; int(0..1) present) { }
foreach ("hello"; int idx; int char) {
  // char is the character CODE (int), e.g. 104 for 'h'
  string ch = sprintf("%c", char);
}

Rule: sprintf and sscanf

sprintf uses Pike format specifiers. %O dumps a readable representation of any value. sscanf returns the number of matched fields, not the parsed values (those are output parameters).

WRONG:

string s = "%d items" % count;              // not Python
array parts = sscanf("42 items", "%d %s");   // wrong: sscanf returns count
write(value);                                 // use write("%O\n", value) to inspect

CORRECT:

string s = sprintf("%d items", count);
int n;
string rest;
int matched = sscanf("42 items", "%d %s", n, rest);
// matched == 2, n == 42, rest == "items"
write("%O\n", some_value); // %O gives readable dump of any type

Rule: zero_type for Missing Keys

m[key] returns 0 (zero) for both a missing key and a key mapped to 0. Use zero_type() to distinguish them.

WRONG:

mapping m = (["a": 0]);
if (!m["a"]) { }          // true — but key EXISTS
if (m["b"] == 0) { }      // true — but key is MISSING
if (m["b"] == UNDEFINED) { } // also true for missing, but zero_type is canonical

CORRECT:

mapping m = (["a": 0]);
if (!zero_type(m["a"])) {
  // key exists (even though value is 0)
}
if (zero_type(m["b"]) == 1) {
  // key is genuinely missing
}
// zero_type returns: 0 (value exists), 1 (UNDEFINED/missing key), 2 (destructed object)

Rule: Operator Overloading via lfuns

Pike operator overloading uses named "lfuns" (low-level functions). These are special method names that Pike calls when operators are used on objects.

WRONG:

class Vec {
  int x, y;
  Vec operator+(Vec other) { } // C++ syntax
  Vec __add__(Vec other) { }   // Python syntax
}

CORRECT:

class Vec {
  int x, y;
  Vec `+(Vec other) {
    return Vec(this->x + other->x, this->y + other->y);
  }
  string _sprintf(int fmt) {
    return sprintf("(%d, %d)", x, y);
  }
  int _equal(mixed other) {
    return objectp(other) && equal(x, other->x) && equal(y, other->y);
  }
}
// Common lfuns: `+, `-, `*, `/, `%, `&, `|, `^, `<<, `>>, `+=, `[..],
//               `==, `<, `>, `!, `[], `[]=, `->, `->=, `(),
//               _sizeof, _values, _indices, _sprintf, _equal,
//               _hash, cast, _random, _search, _types,
//               _serialize, _deserialize, _get_iterator,
//               _m_delete, __hash, _destruct

Additional References

• Syntax Reference — control flow, operators, declarations, preprocessor directives
• Types Reference — type system, coercion, typeof, type annotations
• Standard Library Patterns — 157 sections, 130+ modules, runtime-verified examples
• Idiomatic Pike Guide — autodoc, anti-patterns, data structures, naming conventions