Add np.atleast_1,2,3d() for finfields. Other minor changes.

demos/kmsurvival.py

@@ -77,19 +77,16 @@
 from mpyc.runtime import mpc
 
 
-def fit_plot(T1, T2, E1, E2, title, unit_of_time, label1, label2):
- kmf1 = KaplanMeierFitter()
- kmf2 = KaplanMeierFitter()
- ax = kmf1.fit(T1, E1, label=label1, alpha=0.05).plot(show_censors=True)
- ax = kmf2.fit(T2, E2, label=label2, alpha=0.05).plot(ax=ax, show_censors=True)
+def plot_fits(kmf1, kmf2, title, unit_of_time):
+ ax = kmf1.plot(show_censors=True)
+ ax = kmf2.plot(ax=ax, show_censors=True)
  ax.set_title(title)
  if unit_of_time:
  plt.xlabel(f'timeline ({unit_of_time})')
  lifelines.plotting.add_at_risk_counts(kmf1, kmf2, ax=ax, labels=None)
  plt.tight_layout()
  figname = ax.figure.canvas.manager.get_window_title()
  ax.figure.canvas.manager.set_window_title(f'Party {mpc.pid} - {figname}')
- return kmf1, kmf2
 
 
 def events_to_table(maxT, T, E):
@@ -261,12 +258,15 @@ async def main():
  ' for own events in the clear')
 
  if args.print_tables or args.plot_curves:
- plt.figure(1)
- title = f'Party {mpc.pid}: {name} Survival - individual events'
- kmf1, kmf2 = fit_plot(T1, T2, E1, E2, title, unit_of_time, label1, label2)
+ kmf1 = KaplanMeierFitter(alpha=0.05, label=label1).fit(T1, E1)
+ kmf2 = KaplanMeierFitter(alpha=0.05, label=label2).fit(T2, E2)
  if args.print_tables:
  print(kmf1.event_table)
  print(kmf2.event_table)
+ if args.plot_curves:
+ plt.figure(1)
+ title = f'Party {mpc.pid}: {name} Survival - individual events'
+ plot_fits(kmf1, kmf2, title, unit_of_time)
 
  # expand to timeline 1..maxT and add all input data homomorphically per group
  d1, n1 = events_to_table(maxT, T1, E1)
@@ -286,14 +286,15 @@ async def main():
  ' for aggregated events in the clear')
 
  if args.print_tables or args.plot_curves:
- plt.figure(2)
- title = f'Party {mpc.pid}: {name} Survival - aggregated by {stride} {unit_of_time}'
- kmf1, kmf2 = fit_plot([t * stride for t in T1], [t * stride for t in T2], E1, E2,
- title, unit_of_time, label1, label2)
+ kmf1 = KaplanMeierFitter(alpha=0.05, label=label1).fit([t * stride for t in T1], E1)
+ kmf2 = KaplanMeierFitter(alpha=0.05, label=label2).fit([t * stride for t in T2], E2)
  if args.print_tables:
  print(kmf1.event_table)
  print(kmf2.event_table)
  if args.plot_curves:
+ plt.figure(2)
+ title = f'Party {mpc.pid}: {name} Survival - aggregated by {stride} {unit_of_time}'
+ plot_fits(kmf1, kmf2, title, unit_of_time)
  plt.show()
 
  logging.info('Optimized secure logrank test on all individual events.')

docs/cli.rst

@@ -42,7 +42,7 @@ reserving lower case for the single-letter options of MPyC programs:
 
 .. argparse::
  :module: mpyc
- :func: get_arg_parser
+ :func: _get_arg_parser
  :prog: mpycprog.py
 
  MPyC configuration : @before

docs/mpyc.rst

@@ -4,18 +4,18 @@ MPyC package
 .. automodule:: mpyc
  :members:
 
-mpyc.gmpy
----------
-
-.. automodule:: mpyc.gmpy
- :members:
-
 mpyc.numpy
 ----------
 
 .. automodule:: mpyc.numpy
  :members:
 
+mpyc.gmpy
+---------
+
+.. automodule:: mpyc.gmpy
+ :members:
+
 mpyc.gfpx
 ---------
 

mpyc/README.md

@@ -12,14 +12,14 @@
 8. [sectypes](https://lschoe.github.io/mpyc/mpyc.sectypes.html): SecInt/Fld/Fxp/Flt types for secure (secret-shared) integer/finite-field/fixed-/floating-point values
 9. [runtime](https://lschoe.github.io/mpyc/mpyc.runtime.html): core MPC protocols (many hidden by Python's operator overloading)
 10. [mpctools](https://lschoe.github.io/mpyc/mpyc.mpctools.html): reduce and accumulate with log round complexity
-11. [seclists](https://lschoe.github.io/mpyc/mpyc.seclists.html): secure lists with oblivious access and updates
-12. [secgroups](https://lschoe.github.io/mpyc/mpyc.secgroups.html): SecGrp types for secure (secret-shared) finite group elements
-13. [random](https://lschoe.github.io/mpyc/mpyc.random.html): securely mimicking Python’s [random](https://docs.python.org/3/library/random.html) module
+11. [random](https://lschoe.github.io/mpyc/mpyc.random.html): securely mimicking Python’s [random](https://docs.python.org/3/library/random.html) module
+12. [seclists](https://lschoe.github.io/mpyc/mpyc.seclists.html): secure lists with oblivious access and updates
+13. [secgroups](https://lschoe.github.io/mpyc/mpyc.secgroups.html): SecGrp types for secure (secret-shared) finite group elements
 14. [statistics](https://lschoe.github.io/mpyc/mpyc.statistics.html): securely mimicking Python’s [statistics](https://docs.python.org/3/library/statistics.html) module
 
 The modules are listed in topological order w.r.t. internal dependencies:
 
 - Modules 1-5 are basic modules which can also be used outside an MPC context
 - Modules 6-9 form the core of MPyC
-- Modules 10-12 form the extended core of MPyC
-- Modules 13-14 are small libraries on top of the (extended) core
+- Modules 10-13 form the extended core of MPyC
+- Module 14 is a small library on top of the (extended) core

mpyc/__init__.py

@@ -41,7 +41,7 @@
 import asyncio
 
 
-def get_arg_parser():
+def _get_arg_parser():
  """Return parser for command line arguments passed to the MPyC runtime."""
  parser = argparse.ArgumentParser(add_help=False)
 
@@ -108,7 +108,7 @@ def get_arg_parser():
 
 
 if os.getenv('READTHEDOCS') != 'True':
- options = get_arg_parser().parse_known_args()[0]
+ options = _get_arg_parser().parse_known_args()[0]
  if options.VERSION or options.HELP:
  options.no_log = True
 

mpyc/asyncoro.py

@@ -317,8 +317,7 @@ def __await__(self):
 
 def _nested_list(rt, n, dims):
  if dims:
- n0 = dims[0]
- dims = dims[1:]
+ n0, *dims = dims
  s = [_nested_list(rt, n0, dims) for _ in range(n)]
  else:
  s = [rt() for _ in range(n)]

mpyc/finfields.py

@@ -801,7 +801,11 @@ def __array_function__(self, func, types, args, kwargs):
  a = list(map(cls, a))
  elif isinstance(a, bool) or a is np.True_ or a is np.False_:
  pass
- elif not isinstance(a, tuple): # shape
+ elif func.__name__.startswith('atleast_'):
+ a = tuple(map(lambda _: cls(_, check=False), a))
+ elif isinstance(a, tuple):
+ pass # e.g., for func like shape
+ else:
  a = cls.field(a)
  return a
 
@@ -1342,8 +1346,6 @@ def trace(self, *args, **kwargs):
 
  return type(self)(a, check=True)
 
- # TODO: add atleast1d(a), atleast2d(a), atleast3d(a)
-
  def transpose(self, *axes):
  a = self.value.transpose(*axes)
  return type(self)(a, check=False)

mpyc/runtime.py

@@ -4224,7 +4224,7 @@ def generate_configs(m, addresses):
 
 def setup():
  """Setup a runtime."""
- parser = mpyc.get_arg_parser()
+ parser = mpyc._get_arg_parser()
  argv = sys.argv # keep raw args
  options, args = parser.parse_known_args()
  if options.VERSION:

mpyc/secgroups.py

@@ -733,6 +733,11 @@ def SecGrpFunc(*args, **kwargs):
  globals()[name] = SecGrpFunc
 
 
+SecSymmetricGroup: type
+SecQuadraticResidues: type
+SecSchnorrGroup: type
+SecEllipticCurve: type
+SecClassGroup: type
 _toSecGrpFunc(fg.SymmetricGroup) # make SecSymmetricGroup as secure SymmetricGroup version
 _toSecGrpFunc(fg.QuadraticResidues) # make SecQuadraticResidues as secure QuadraticResidues version
 _toSecGrpFunc(fg.SchnorrGroup) # make SecSchnorrGroup as secure SchnorrGroup version

mpyc/sectypes.py

@@ -1009,7 +1009,8 @@ def __bool__(self):
  def __array_function__(self, func, types, args, kwargs):
  # minimal redirect for now
  if f'{func.__name__}' == 'vstack':
- # NB: Numpy 2.0 inserts keyword arguments for np.vstack when converting deprecated np.row_stack call.
+ # NB: Numpy 2.0 inserts keyword arguments for np.vstack
+ # when converting deprecated np.row_stack call.
  kwargs = {}
  return eval(f'runtime.np_{func.__name__}')(*args, **kwargs)
 

mpyc/statistics.py

@@ -602,11 +602,11 @@ def linear_regression(x, y):
  """
  n = len(x)
  if len(y) != n:
- raise statistics.StatisticsError('covariance requires that both inputs '
+ raise statistics.StatisticsError('linear regression requires that both inputs '
  'have same number of data points')
 
  if n < 2:
- raise statistics.StatisticsError('covariance requires at least two data points')
+ raise statistics.StatisticsError('linear regression requires at least two data points')
 
  sectype = type(x[0]) # all elts of x assumed of same type
  if not issubclass(sectype, SecureObject):

tests/test_finfields.py

@@ -468,6 +468,12 @@ def test_array_function(self):
  np.assertEqual(np.tensordot(F_a, F_b), np.tensordot(a, b))
  np.assertEqual(np.inner(F_a, F_b), np.inner(a, b))
  np.assertEqual(np.outer(F_a, F_b), np.outer(a, b))
+ np.assertEqual(np.atleast_1d(F(1)), np.atleast_1d(1))
+ np.assertEqual(np.atleast_1d(F(1), F_a)[1], np.atleast_1d(1, a)[1])
+ np.assertEqual(np.atleast_2d(F(1)), np.atleast_2d(1))
+ np.assertEqual(np.atleast_2d(F(1), F_a)[1], np.atleast_2d(1, a)[1])
+ np.assertEqual(np.atleast_3d(F(1)), np.atleast_3d(1))
+ np.assertEqual(np.atleast_3d(F(1), F_a)[1], np.atleast_3d(1, a)[1])
  np.assertEqual(np.concatenate((F_a, F_b, F_a.T)), np.concatenate((a, b, a.T)))
  np.assertEqual(np.stack([F_a, F_b]), np.stack([a, b]))
  np.assertEqual(np.stack([a, F_b], axis=1), np.stack([a, b], axis=1))
@@ -478,7 +484,8 @@ def test_array_function(self):
  np.assertEqual(np.hstack([F_a, F_b]), np.hstack([a, b]))
  np.assertEqual(np.dstack([F_a, F_b]), np.dstack([a, b]))
  np.assertEqual(np.column_stack([F_a, F_b]), np.column_stack([a, b]))
- np.assertEqual(np.row_stack([F_a, F_b]), np.row_stack([a, b]))
+ if np.lib.NumpyVersion(np.__version__) < '2.0.0b1':
+ np.assertEqual(np.row_stack([F_a, F_b]), np.row_stack([a, b]))
  np.assertEqual(np.split(F_a, 2)[0], np.split(a, 2)[0])
  np.assertEqual(np.array_split(F_a, 2)[1], np.array_split(a, 2)[1])
  np.assertEqual(np.dsplit(F_a.reshape(1, 4, 4), 2)[1], np.dsplit(a.reshape(1, 4, 4), 2)[1])

tests/test_runtime.py

@@ -59,7 +59,8 @@ def test_secint_array(self):
  np.assertEqual(mpc.run(mpc.output(np.dstack((c[0], c[0])))), np.dstack((a[0], a[0])))
  np.assertEqual(mpc.run(mpc.output(np.dstack((c[0, 0],)))), np.dstack((a[0, 0],)))
  np.assertEqual(mpc.run(mpc.output(np.column_stack((c, c, c)))), np.column_stack((a, a, a)))
- np.assertEqual(mpc.run(mpc.output(np.row_stack((c, c, c)))), np.row_stack((a, a, a)))
+ if np.lib.NumpyVersion(np.__version__) < '2.0.0b1':
+ np.assertEqual(mpc.run(mpc.output(np.row_stack((c, c, c)))), np.row_stack((a, a, a)))
  np.assertEqual(mpc.run(mpc.output(np.split(c, 2, 1)[0])), np.split(a, 2, 1)[0])
  np.assertEqual(mpc.run(mpc.output(np.dsplit(d, 1)[0])), np.dsplit(b, 1)[0])
  np.assertEqual(mpc.run(mpc.output(np.hsplit(c, 2)[0])), np.hsplit(a, 2)[0])