more common/ pruning (#30778)

common/ffi_wrapper.py

@@ -1,55 +1,8 @@
-import os
-import sys
-import fcntl
-import hashlib
 import platform
-from cffi import FFI
+
 
 def suffix():
   if platform.system() == "Darwin":
     return ".dylib"
   else:
     return ".so"
-
-def ffi_wrap(name, c_code, c_header, tmpdir="/tmp/ccache", cflags="", libraries=None):
-  if libraries is None:
-    libraries = []
-
-  cache = name + "_" + hashlib.sha1(c_code.encode('utf-8')).hexdigest()
-  try:
-    os.mkdir(tmpdir)
-  except OSError:
-    pass
-
-  fd = os.open(tmpdir, 0)
-  fcntl.flock(fd, fcntl.LOCK_EX)
-  try:
-    sys.path.append(tmpdir)
-    try:
-      mod = __import__(cache)
-    except Exception:
-      print(f"cache miss {cache}")
-      compile_code(cache, c_code, c_header, tmpdir, cflags, libraries)
-      mod = __import__(cache)
-  finally:
-    os.close(fd)
-
-  return mod.ffi, mod.lib
-
-
-def compile_code(name, c_code, c_header, directory, cflags="", libraries=None):
-  if libraries is None:
-    libraries = []
-
-  ffibuilder = FFI()
-  ffibuilder.set_source(name, c_code, source_extension='.cpp', libraries=libraries)
-  ffibuilder.cdef(c_header)
-  os.environ['OPT'] = "-fwrapv -O2 -DNDEBUG -std=c++1z"
-  os.environ['CFLAGS'] = cflags
-  ffibuilder.compile(verbose=True, debug=False, tmpdir=directory)
-
-
-def wrap_compiled(name, directory):
-  sys.path.append(directory)
-  mod = __import__(name)
-  return mod.ffi, mod.lib

common/lazy_property.py

@@ -1,12 +0,0 @@
-class lazy_property():
-  """Defines a property whose value will be computed only once and as needed.
-
-     This can only be used on instance methods.
-  """
-  def __init__(self, func):
-    self._func = func
-
-  def __get__(self, obj_self, cls):
-    value = self._func(obj_self)
-    setattr(obj_self, self._func.__name__, value)
-    return value

common/numpy_helpers.py

@@ -1,22 +0,0 @@
-import numpy as np
-
-
-def deep_interp_np(x, xp, fp, axis=None):
-  if axis is not None:
-    fp = fp.swapaxes(0,axis)
-  x = np.atleast_1d(x)
-  xp = np.array(xp)
-  if len(xp) < 2:
-    return np.repeat(fp, len(x), axis=0)
-  if min(np.diff(xp)) < 0:
-    raise RuntimeError('Bad x array for interpolation')
-  j = np.searchsorted(xp, x) - 1
-  j = np.clip(j, 0, len(xp)-2)
-  d = np.divide(x - xp[j], xp[j + 1] - xp[j], out=np.ones_like(x, dtype=np.float64), where=xp[j + 1] - xp[j] != 0)
-  vals_interp = (fp[j].T*(1 - d)).T + (fp[j + 1].T*d).T
-  if axis is not None:
-    vals_interp = vals_interp.swapaxes(0,axis)
-  if len(vals_interp) == 1:
-    return vals_interp[0]
-  else:
-    return vals_interp