Class: String

Inherits:

Object

• Object
• String

Defined in:

lib/sin_fast_blank.rb

Instance Method Summary

• #ascii_blank? ⇒ Boolean
• #blank? ⇒ Boolean

Instance Method Details

#ascii_blank? ⇒ Boolean

Returns:

• (Boolean)

# File 'ext/sin_fast_blank/sin_fast_blank.c', line 445

static VALUE rb_str_ascii_blank(VALUE str) {
  long len = RSTRING_LEN(str);
  if (len == 0) return Qtrue;

  const unsigned char* ptr = (const unsigned char*)RSTRING_PTR(str);
  const unsigned char* end = ptr + len;
  rb_encoding* enc = STR_ENC_GET(str);

  /*
   * This one never raises, because bytes that decode to nothing are no more an ASCII blank than the characters they failed to form.
   * An ASCII-compatible encoding does not even need the decoder: a byte of 0x80 or above only ever starts a character whose codepoint
   * is 0x80 or above there, so the first one settles the answer.
   */
  if (rb_enc_asciicompat(enc)) return check_ascii_blank(ptr, (size_t)len) ? Qtrue : Qfalse;

  while (ptr < end) {
    int clen = rb_enc_precise_mbclen((const char*)ptr, (const char*)end, enc);
    if (!MBCLEN_CHARFOUND_P(clen)) return Qfalse;
    unsigned int codepoint = rb_enc_mbc_to_codepoint((const char*)ptr, (const char*)end, enc);
    if (codepoint != 0 && !rb_isspace(codepoint)) return Qfalse;
    ptr += MBCLEN_CHARFOUND_LEN(clen);
  }

  return Qtrue;
}

#blank? ⇒ Boolean

Returns:

• (Boolean)

# File 'ext/sin_fast_blank/sin_fast_blank.c', line 398

static VALUE rb_str_blank(VALUE str) {
  long len = RSTRING_LEN(str);
  if (len == 0) return Qtrue;

  const unsigned char* ptr = (const unsigned char*)RSTRING_PTR(str);
  const unsigned char* end = ptr + len;
  rb_encoding* enc = STR_ENC_GET(str);

  bool asciicompat = rb_enc_asciicompat(enc) != 0;
  if (asciicompat) {
    const unsigned char* non_ascii_pos = NULL;

    if (check_blank(ptr, (size_t)len, &non_ascii_pos)) return Qtrue;
    if (non_ascii_pos == NULL) return Qfalse;

    ptr = non_ascii_pos;
  }

  bool is_unicode = is_unicode_encoding(enc);
  while (ptr < end) {
    int clen = rb_enc_precise_mbclen((const char*)ptr, (const char*)end, enc);
    if (!MBCLEN_CHARFOUND_P(clen)) return blank_undecodable(str, enc);
    unsigned int codepoint = rb_enc_mbc_to_codepoint((const char*)ptr, (const char*)end, enc);
    if (!is_blank_codepoint(codepoint, enc, is_unicode)) return Qfalse;
    ptr += MBCLEN_CHARFOUND_LEN(clen);

    /*
     * An ASCII run starts here, so hand it back to the SIMD scan instead of decoding it a character at a time. Only an ASCII-compatible
     * encoding may do this: anywhere else a byte below 0x80 is not a character on its own.
     *
     * Resuming the decode afterwards is safe too. The scan only ever hands back a position holding a byte of 0x80 or above, and every
     * byte it passed was a single-byte blank, so the decode restarts on a character boundary. A non-blank ASCII byte settles the answer
     * outright and leaves no position to hand back.
     */
    if (asciicompat && ptr < end && *ptr < 0x80) {
      const unsigned char* non_ascii_pos = NULL;

      if (check_blank(ptr, (size_t)(end - ptr), &non_ascii_pos)) return Qtrue;
      if (non_ascii_pos == NULL) return Qfalse;

      ptr = non_ascii_pos;
    }
  }

  return Qtrue;
}