mycpp

#include <errno.h>  // errno

#include <float.h>  // DBL_MIN, DBL_MAX

#include <math.h>   // INFINITY

#include <stdio.h>  // required for readline/readline.h (man readline)

#include "_build/detected-cpp-config.h"

#include "mycpp/gc_list.h"

#include "mycpp/gc_str.h"

#ifdef HAVE_READLINE

  #include "cpp/frontend_pyreadline.h"

#endif

// Translation of Python's print().

void print(BigStr* s) {

  fputs(s->data_, stdout);  // print until first NUL

  fputc('\n', stdout);

}

BigStr* str(int i) {

  BigStr* s = OverAllocatedStr(kIntBufSize);

  int length = snprintf(s->data(), kIntBufSize, "%d", i);

  s->MaybeShrink(length);

  return s;

}

BigStr* str(double d) {

  char buf[64];  // overestimate, but we use snprintf() to be safe

  int n = sizeof(buf) - 2;  // in case we add '.0'

  // The round tripping test in mycpp/float_test.cc tells us:

  // %.9g - FLOAT round trip

  // %.17g - DOUBLE round trip

  // But this causes problems in practice, e.g. for 3.14, or 1/3

  // int length = snprintf(buf, n, "%.17g", d);

  // So use 1 less digit, which happens to match Python 3 and node.js (but not

  // Python 2)

  int length = snprintf(buf, n, "%.16g", d);

  // TODO: This may depend on LC_NUMERIC locale!

  // We may return the strings:

  //    inf  -inf   nan

  // But this shouldn't come up much, because Python code changes it to:

  //    INFINITY   -INFINITY   NAN

  if (strchr(buf, 'i') || strchr(buf, 'n')) {

    return StrFromC(buf);  // don't add .0

  }

  // Problem:

  // %f prints 3.0000000 and 3.500000

  // %g prints 3 and 3.5

  //

  // We want 3.0 and 3.5, so add '.0' in some cases

  if (!strchr(buf, '.')) {  // 12345 -> 12345.0

    buf[length] = '.';

    buf[length + 1] = '0';

    buf[length + 2] = '\0';

  }

  return StrFromC(buf);

}

// %a is a hexfloat form, probably don't need that

// int length = snprintf(buf, n, "%a", d);

// Do we need this API?  Or is mylib.InternedStr(BigStr* s, int start, int end)

// better for getting values out of Token.line without allocating?

//

// e.g. mylib.InternedStr(tok.line, tok.start, tok.start+1)

//

// Also for SmallStr, we don't care about interning.  Only for HeapStr.

BigStr* intern(BigStr* s) {

  // TODO: put in table gHeap.interned_

  return s;

}

// Print quoted string.  Called by StrFormat('%r').

// TODO: consider using J8 notation instead, since error messages show that

// string.

BigStr* repr(BigStr* s) {

  // Worst case: \0 becomes 4 bytes as '\\x00', and then two quote bytes.

  int n = len(s);

  int upper_bound = n * 4 + 2;

  BigStr* result = OverAllocatedStr(upper_bound);

  // Single quote by default.

  char quote = '\'';

  if (memchr(s->data_, '\'', n) && !memchr(s->data_, '"', n)) {

    quote = '"';

  }

  char* p = result->data_;

  // From PyString_Repr()

  *p++ = quote;

  for (int i = 0; i < n; ++i) {

    unsigned char c = static_cast<unsigned char>(s->data_[i]);

    if (c == quote || c == '\\') {

      *p++ = '\\';

      *p++ = c;

    } else if (c == '\t') {

      *p++ = '\\';

      *p++ = 't';

    } else if (c == '\n') {

      *p++ = '\\';

      *p++ = 'n';

    } else if (c == '\r') {

      *p++ = '\\';

      *p++ = 'r';

    } else if (0x20 <= c && c < 0x80) {

      *p++ = c;

    } else {

      // Unprintable becomes \xff.

      // TODO: Consider \yff.  This is similar to J8 strings, but we don't

      // decode UTF-8.

      sprintf(p, "\\x%02x", c & 0xff);

      p += 4;

    }

  }

  *p++ = quote;

  *p = '\0';

  int length = p - result->data_;

  result->MaybeShrink(length);

  return result;

}

// Helper functions that don't use exceptions.

bool StringToInt(const char* s, int length, int base, int* result) {

  if (length == 0) {

    return false;  // empty string isn't a valid integer

  }

  // Note: sizeof(int) is often 4 bytes on both 32-bit and 64-bit

  //       sizeof(long) is often 4 bytes on both 32-bit but 8 bytes on 64-bit

  // static_assert(sizeof(long) == 8);

  char* pos;  // mutated by strtol

  errno = 0;

  long v = strtol(s, &pos, base);

  if (errno == ERANGE) {

    switch (v) {

    case LONG_MIN:

      return false;  // underflow of long, which may be 64 bits

    case LONG_MAX:

      return false;  // overflow of long

    }

  }

  // It should ALSO fit in an int, not just a long

  if (v > INT_MAX) {

    return false;

  }

  if (v < INT_MIN) {

    return false;

  }

  const char* end = s + length;

  if (pos == end) {

    *result = v;

    return true;  // strtol() consumed ALL characters.

  }

  while (pos < end) {

    if (!IsAsciiWhitespace(*pos)) {

      return false;  // Trailing non-space

    }

    pos++;

  }

  *result = v;

  return true;  // Trailing space is OK

}

bool StringToInt64(const char* s, int length, int base, int64_t* result) {

  if (length == 0) {

    return false;  // empty string isn't a valid integer

  }

  // These should be the same type

  static_assert(sizeof(long long) == sizeof(int64_t), "");

  char* pos;  // mutated by strtol

  errno = 0;

  long long v = strtoll(s, &pos, base);

  if (errno == ERANGE) {

    switch (v) {

    case LLONG_MIN:

      return false;  // underflow

    case LLONG_MAX:

      return false;  // overflow

    }

  }

  const char* end = s + length;

  if (pos == end) {

    *result = v;

    return true;  // strtol() consumed ALL characters.

  }

  while (pos < end) {

    if (!IsAsciiWhitespace(*pos)) {

      return false;  // Trailing non-space

    }

    pos++;

  }

  *result = v;

  return true;  // Trailing space is OK

}

int to_int(BigStr* s, int base) {

  int i;

  if (StringToInt(s->data_, len(s), base, &i)) {

    return i;  // truncated to int

  } else {

    throw Alloc<ValueError>();

  }

}

BigStr* chr(int i) {

  // NOTE: i should be less than 256, in which we could return an object from

  // GLOBAL_STR() pool, like StrIter

  auto result = NewStr(1);

  result->data_[0] = i;

  return result;

}

int ord(BigStr* s) {

  assert(len(s) == 1);

  // signed to unsigned conversion, so we don't get values like -127

  uint8_t c = static_cast<uint8_t>(s->data_[0]);

  return c;

}

bool to_bool(BigStr* s) {

  return len(s) != 0;

}

double to_float(int i) {

  return static_cast<double>(i);

}

double to_float(BigStr* s) {

  char* begin = s->data_;

  char* end_pos = begin + len(s);

  char* orig_end = end_pos;

  errno = 0;

  double result = strtod(begin, &end_pos);

  if (errno == ERANGE) {  // error: overflow or underflow

    if (result >= HUGE_VAL) {

      return INFINITY;

    } else if (result <= -HUGE_VAL) {

      return -INFINITY;

    } else if (-DBL_MIN <= result && result <= DBL_MIN) {

      return 0.0;

    } else {

      FAIL("Invalid value after ERANGE");

    }

  }

  if (end_pos == begin) {  // error: not a floating point number

    throw Alloc<ValueError>();

  }

  if (end_pos != orig_end) {  // trailing data like '5s' not alowed

    while (end_pos < orig_end) {

      if (!IsAsciiWhitespace(*end_pos)) {

        throw Alloc<ValueError>();  // Trailing non-space

      }

      end_pos++;

    }

  }

  return result;

}

// e.g. ('a' in 'abc')

bool str_contains(BigStr* haystack, BigStr* needle) {

  // Common case

  if (len(needle) == 1) {

    return memchr(haystack->data_, needle->data_[0], len(haystack));

  }

  if (len(needle) > len(haystack)) {

    return false;

  }

  // General case. TODO: We could use a smarter substring algorithm.

  const char* end = haystack->data_ + len(haystack);

  const char* last_possible = end - len(needle);

  const char* p = haystack->data_;

  while (p <= last_possible) {

    if (memcmp(p, needle->data_, len(needle)) == 0) {

      return true;

    }

    p++;

  }

  return false;

}

BigStr* str_repeat(BigStr* s, int times) {

  // Python allows -1 too, and Oil used that

  if (times <= 0) {

    return kEmptyString;

  }

  int len_ = len(s);

  int new_len = len_ * times;

  BigStr* result = NewStr(new_len);

  char* dest = result->data_;

  for (int i = 0; i < times; i++) {

    memcpy(dest, s->data_, len_);

    dest += len_;

  }

  return result;

}

// for os_path.join()

// NOTE(Jesse): Perfect candidate for BoundedBuffer

BigStr* str_concat3(BigStr* a, BigStr* b, BigStr* c) {

  int a_len = len(a);

  int b_len = len(b);

  int c_len = len(c);

  int new_len = a_len + b_len + c_len;

  BigStr* result = NewStr(new_len);

  char* pos = result->data_;

  memcpy(pos, a->data_, a_len);

  pos += a_len;

  memcpy(pos, b->data_, b_len);

  pos += b_len;

  memcpy(pos, c->data_, c_len);

  assert(pos + c_len == result->data_ + new_len);

  return result;

}

BigStr* str_concat(BigStr* a, BigStr* b) {

  int a_len = len(a);

  int b_len = len(b);

  int new_len = a_len + b_len;

  BigStr* result = NewStr(new_len);

  char* buf = result->data_;

  memcpy(buf, a->data_, a_len);

  memcpy(buf + a_len, b->data_, b_len);

  return result;

}

//

// Comparators

//

bool str_equals(BigStr* left, BigStr* right) {

  // Fast path for identical strings.  String deduplication during GC could

  // make this more likely.  String interning could guarantee it, allowing us

  // to remove memcmp().

  if (left == right) {

    return true;

  }

  // TODO: It would be nice to remove this condition, but I think we need MyPy

  // strict None checking for it

  if (left == nullptr || right == nullptr) {

    return false;

  }

  if (left->len_ != right->len_) {

    return false;

  }

  return memcmp(left->data_, right->data_, left->len_) == 0;

}

bool maybe_str_equals(BigStr* left, BigStr* right) {

  if (left && right) {

    return str_equals(left, right);

  }

  if (!left && !right) {

    return true;  // None == None

  }

  return false;  // one is None and one is a BigStr*

}

bool items_equal(BigStr* left, BigStr* right) {

  return str_equals(left, right);

}

bool keys_equal(BigStr* left, BigStr* right) {

  return items_equal(left, right);

}

bool items_equal(Tuple2<int, int>* t1, Tuple2<int, int>* t2) {

  return (t1->at0() == t2->at0()) && (t1->at1() == t2->at1());

}

bool keys_equal(Tuple2<int, int>* t1, Tuple2<int, int>* t2) {

  return items_equal(t1, t2);

}

bool items_equal(Tuple2<BigStr*, int>* t1, Tuple2<BigStr*, int>* t2) {

  return items_equal(t1->at0(), t2->at0()) && (t1->at1() == t2->at1());

}

bool keys_equal(Tuple2<BigStr*, int>* t1, Tuple2<BigStr*, int>* t2) {

  return items_equal(t1, t2);

}

bool str_equals_c(BigStr* s, const char* c_string, int c_len) {

  // Needs SmallStr change

  if (len(s) == c_len) {

    return memcmp(s->data_, c_string, c_len) == 0;

  } else {

    return false;

  }

}

bool str_equals0(const char* c_string, BigStr* s) {

  int n = strlen(c_string);

  if (len(s) == n) {

    return memcmp(s->data_, c_string, n) == 0;

  } else {

    return false;

  }

}

int hash(BigStr* s) {

  return s->hash(fnv1);

}

int max(int a, int b) {

  return std::max(a, b);

}

int min(int a, int b) {

  return std::min(a, b);

}

int max(List<int>* elems) {

  int n = len(elems);

  if (n < 1) {

    throw Alloc<ValueError>();

  }

  int ret = elems->at(0);

  for (int i = 0; i < n; ++i) {

    int cand = elems->at(i);

    if (cand > ret) {

      ret = cand;

    }

  }

  return ret;

}