1. 15 Aug, 2008 7 commits
    • Dag Sverre Seljebotn's avatar
    • Dag Sverre Seljebotn's avatar
      numpy.pxd improvements, see details · 62ef7487
      Dag Sverre Seljebotn authored
      Do not require malloc for format string for common cases.
      Provide __cythonbufferdefaults__
      More robust for changing NumPy APIs
      ;
      62ef7487
    • Dag Sverre Seljebotn's avatar
      Make numpy tests run on my doctest · a4572ec9
      Dag Sverre Seljebotn authored
      a4572ec9
    • JimKleckner's avatar
      Windos __stdcall and _USE_MATH_DEFINES · 9035ac46
      JimKleckner authored
      9035ac46
    • Robert Bradshaw's avatar
      More NumPy array fixes. · d90d2455
      Robert Bradshaw authored
      d90d2455
    • Stefan Behnel's avatar
      removed debug code · 43bef567
      Stefan Behnel authored
      43bef567
    • Stefan Behnel's avatar
      Rewrite of the string literal handling code · 2e8a0084
      Stefan Behnel authored
      String literals pass through the compiler as follows:
      - unicode string literals are stored as unicode strings and encoded to UTF-8 on the way out
      - byte string literals are stored as correctly encoded byte strings by unescaping the source string literal into the corresponding byte sequence. No further encoding is done later on!
      - char literals are stored as byte strings of length 1. This can be verified by the parser now, e.g. a non-ASCII char literal in UTF-8 source code will result in an error, as it would end up as two or more bytes in the C code, which can no longer be represented as a C char.
      
      Storing byte strings is necessary as we otherwise loose the ability to encode byte string literals on the way out. They do not necessarily contain only bytes that fit into the source code encoding as the source can use escape sequences to represent them. Previously, ASCII encoded source code could not contain byte string literals with properly escaped non-ASCII bytes.
      
      Another bug that was fixed: in Python, escape sequences behave different in unicode strings (where they represent the character code) and byte strings (where they represent a byte value). Previously, they resulted in the same byte value in Cython code. This is only a problem for non-ASCII escapes, since the character code and the byte value of ASCII characters are identical.
      2e8a0084
  2. 14 Aug, 2008 6 commits
  3. 13 Aug, 2008 12 commits
  4. 12 Aug, 2008 11 commits
  5. 11 Aug, 2008 4 commits