snappy 3.2__cp310-cp310-macosx_10_12_x86_64.whl

snappy/numeric_output_checker.py

@@ -0,0 +1,345 @@
+"""
+Provides the NumericOutputChecker implementing doctest's OutputChecker.
+
+It provides a set of extra flags NUMERIC6, NUMERIC9, ... for doctests
+allowing numbers in an example to differ by an epsilon = 1e-6, 1e-9, ...
+
+See init_precisions(...) for the precisions we support.
+
+See documentation of NumericExample for more.
+"""
+
+
+class NumericExample:
+    """
+    The actual result (3.141592653589793) differs from the given result,
+    but the difference is less than 1e-6, so it still passes::
+
+        >>> 1.5707963267948966 * 2 # doctest: +NUMERIC6
+        3.14159285
+
+    The text pieces between the numbers are also compared, performing
+    white-space normalization::
+
+        >>> ['a', 4.5, 6.9] # doctest: +NUMERIC6
+        ['a', 4.5,       6.9000000001]
+
+    Intervals in the notation emitted by sage are allowed::
+
+        >>> print("4.5?e-1") # doctest: +NUMERIC6
+        4.50000000001?e-1
+
+    """
+
+
+import doctest
+import re
+import decimal
+
+# Generate flags NUMERIC6, ...
+
+# Store as pairs (precision, FLAG)
+NUMERIC_LIST = []
+# Store in dict precision : FLAG
+NUMERIC_DICT = {}
+# All or'ed together
+ALL_NUMERIC = 0
+
+
+def init_precisions(precisions):
+    """
+    Register flags for given precisions with doctest module.
+
+    Unfortunately, doctest doesn't seem to support a more generic mechanism
+    such as "# doctest: +NUMERIC: 6" to specify the precision and we need to
+    unroll each precision we want to its own flag.
+    """
+
+    global NUMERIC_LIST
+    global NUMERIC_DICT
+    global ALL_NUMERIC
+
+    for precision in precisions:
+        # Check key such that clients having demand for different
+        # precisions could register them.
+        if precision not in NUMERIC_DICT:
+            flag = doctest.register_optionflag('NUMERIC%d' % precision)
+            NUMERIC_LIST.append((precision, flag))
+            NUMERIC_DICT[precision] = flag
+            ALL_NUMERIC |= flag
+
+
+# The precisions NUMERIC0, ... we support are hard-coded here:
+init_precisions(range(0,33,3))
+
+
+def get_precision(optionflags):
+    """
+    Get precision from optionflags
+    """
+    for precision, flag in NUMERIC_LIST:
+        if optionflags & flag:
+            return precision
+
+# Regular expressions matching numbers and intervals such as
+# 23, 2.3, 3.5e-4, 3.4 E-5, 3.4?, 5.6?e-3
+
+
+mantissa_pat = r'(-?[0-9]+(?:\.[0-9]+)?)'
+# Intervals can be 3.4? or 3.4?e-2, account for the "?"
+interval_pat = r'(\?)?'
+# Pari prints 3.4E-3 as "3.4 E-3", account for the space
+exponent_pat = r'(?:\ ?([eE][+-]?[0-9]+))?'
+
+# 4.5?e-3 would split into groups ('4.5?e-3', '4.5', '?", 'e-3').
+number_re = re.compile('(' + mantissa_pat + interval_pat + exponent_pat + ')')
+number_group_count = 4
+number_split_stride = number_group_count + 1
+
+# Use whitespace normalization for text pieces
+NUMERIC_DEFAULT_OPTIONFLAGS = doctest.NORMALIZE_WHITESPACE
+
+# Given the above groups, e.g., ('4.5?e-3', '4.5', '?", 'e-3'), return
+# decimal.Decimal("4.5E-3")
+
+
+def to_decimal(groups):
+    number, mantissa, interval, exponent = groups
+    if exponent:
+        n = mantissa + exponent
+    else:
+        n = mantissa
+    return decimal.Decimal(n)
+
+
+class NumericOutputChecker(doctest.OutputChecker):
+    """
+    Implements doctest's OutputChecker, see documentation of
+    NumericExample for examples.
+
+    >>> N = NumericOutputChecker()
+
+    >>> a   = "[3.499e-8, 4.5?e-8]"
+    >>> b   = "[3.499e-8,   4.5?e-8]"
+
+    >>> N.check_output(a, b, NUMERIC_DICT[12])
+    True
+
+    >>> b   = "[3.499999e-8,   3.2?e-8]"
+    >>> N.check_output(a, b, NUMERIC_DICT[6])
+    True
+    >>> N.check_output(a, b, NUMERIC_DICT[9])
+    False
+    >>> N.formatted_compare_numeric(a, b, NUMERIC_DICT[9])
+    'Numbers differed by 1.3E-8\\n\\nExpected     : 3.499e-8\\nGot          : 3.499999e-8\\nDifference          : 9.99E-12\\n\\nExpected     : 4.5?e-8\\nGot          : 3.2?e-8\\nDifference (FAILURE): 1.3E-8\\n'
+    >>> N.compare_numeric(a, b, NUMERIC_DICT[12])
+    ('NUMERIC', ([('3.499e-8', '3.499999e-8', True, Decimal('9.99E-12')), ('4.5?e-8', '3.2?e-8', True, Decimal('1.3E-8'))], Decimal('1.3E-8')))
+
+    >>> b   = "[3.4999e-8,  4.5e-8]"
+    >>> N.formatted_compare_numeric(a, b, NUMERIC_DICT[6])
+    'Expected interval, but got 4.5e-8.'
+
+    >>> b   = "[3.4999?e-8, 4.5e-8]"
+    >>> N.formatted_compare_numeric(a, b, NUMERIC_DICT[6])
+    'Expected number, but got 3.4999?e-8.'
+
+    >>> b  = "a = [3.4999e-8,  4.5?e-8]"
+    >>> N.formatted_compare_numeric(a, b, NUMERIC_DICT[6])
+    'Text between numbers differs: Expected "[" but got "a = [" at position 0'
+
+    >>> b  = "[3.4999e-8,  4.5?e-8, 5.63]"
+    >>> N.formatted_compare_numeric(a, b, NUMERIC_DICT[6])
+    'Expected 2 numbers but got 3 numbers.'
+
+    >>> a   = "[4.5,       6.7e1,       2e+3]"
+    >>> b   = "[4.5000001, 67.00000001, 2.0000000000000000001e+3]"
+    >>> N.compare_numeric(a, b, NUMERIC_DICT[6])
+    ('OK', None)
+    >>> N.compare_numeric(a, b, NUMERIC_DICT[12])
+    ('NUMERIC', ([('4.5', '4.5000001', True, Decimal('1E-7')), ('6.7e1', '67.00000001', True, Decimal('1E-8')), ('2e+3', '2.0000000000000000001e+3', False, Decimal('1E-16'))], Decimal('1E-7')))
+
+    Account for pari adding a space before the E::
+
+        >>> a   = "4.5e-9"
+        >>> b   = "4.5 E-9"
+        >>> N.compare_numeric(a, b, NUMERIC_DICT[12])
+        ('OK', None)
+
+    """
+
+    def compare_numeric(self, want, got, optionflags):
+        """
+        Compares want and got by scanning for numbers. The numbers are
+        compared using an epsilon extracted from optionflags. The text
+        pieces between the numbers are compared falling back to the
+        default implementation of OutputChecker.
+
+        Returns a pair (status, data) where status is 'OK' if the
+        comparison passed or indicates how it failed with data containing
+        information that can be used to format the text explaining the
+        differences.
+        """
+
+        # "[4.5e-9]" yields ('[', '4.5e-9', '4.5', None, 'e-9', ']')
+        split_want = re.split(number_re, want)
+        split_got = re.split(number_re, got)
+
+        # Check same number of numbers
+        if len(split_want) != len(split_got):
+            return ('COUNT',
+                    (len(split_want) // number_split_stride,
+                     len(split_got) // number_split_stride))
+
+        # Compare text pieces between numbers
+        flags = optionflags | NUMERIC_DEFAULT_OPTIONFLAGS
+
+        for i in range(0, len(split_want), number_split_stride):
+            if not doctest.OutputChecker.check_output(
+                    self, split_want[i], split_got[i], flags):
+                return ('TEXT', (split_want[i], split_got[i], i))
+
+        epsilon = decimal.Decimal(0.1) ** get_precision(optionflags)
+
+        rows = []
+        max_diff = 0
+
+        for i in range(1, len(split_want), number_split_stride):
+            # Number how it literally appears
+            number_want = split_want[i]
+            number_got = split_got[i]
+
+            # Check whether both intervals or numbers
+            is_interval_want = bool(split_want[i + 2])
+            is_interval_got = bool(split_got[i + 2])
+            if is_interval_want != is_interval_got:
+                return ('TYPE', (is_interval_want, number_got))
+
+            # Number (or crushed interval) as decimal.Decimal
+            decimal_want = to_decimal(split_want[i: i + number_group_count])
+            decimal_got = to_decimal(split_got[i: i + number_group_count])
+
+            # Compute diff
+            diff = abs(decimal_want - decimal_got)
+            failed = (diff > epsilon)
+
+            max_diff = max(max_diff, diff)
+
+            # Record data
+            rows.append((number_want, number_got, failed, diff))
+
+        if max_diff > epsilon:
+            # Larger than epsilon: failed
+            return ('NUMERIC', (rows, max_diff))
+
+        # Ok
+        return ('OK', None)
+
+    def format_compare_numeric_result(self, status, data):
+        """
+        Formats a nice text from the result of compare_numeric.
+        """
+
+        if status == 'COUNT':
+            return 'Expected %d numbers but got %d numbers.' % data
+        elif status == 'TEXT':
+            return 'Text between numbers differs: Expected "%s" but got "%s" at position %d' % data
+        elif status == 'TYPE':
+            is_interval_want, number_got = data
+            if is_interval_want:
+                k = 'interval'
+            else:
+                k = 'number'
+            return 'Expected %s, but got %s.' % (k, number_got)
+        elif status == 'NUMERIC':
+            rows, max_diff = data
+            result = 'Numbers differed by %s\n' % max_diff
+            for number_want, number_got, failed, diff in rows:
+                if result:
+                    result += '\n'
+                result += 'Expected     : %s\n' % number_want
+                result += 'Got          : %s\n' % number_got
+                if failed:
+                    result += 'Difference (FAILURE): %s\n' % diff
+                else:
+                    result += 'Difference          : %s\n' % diff
+            return result
+        else:
+            raise Exception('Internal error in OutputChecker.')
+
+    def formatted_compare_numeric(self, want, got, optionflags):
+        """
+        Performs comparison of compare_numeric and returns formatted
+        text.
+
+        Only supposed to be used if comparison failed.
+        """
+
+        status, data = self.compare_numeric(want, got, optionflags)
+
+        return self.format_compare_numeric_result(status, data)
+
+    def check_output(self, want, got, optionflags):
+        """
+        Implementation of OutputChecker method.
+        """
+
+        if want == got:
+            # Early bail
+            return True
+
+        if optionflags & ALL_NUMERIC:
+            # Use our method if any NUMERIC flag set
+            status, data = self.compare_numeric(want, got, optionflags)
+            return status == 'OK'
+        else:
+            # Otherwise fallback to base implementation.
+            return doctest.OutputChecker.check_output(
+                self, want, got, optionflags)
+
+    def output_difference(self, example, got, optionflags):
+        """
+        Implementation of OutputChecker method.
+        """
+        if (not optionflags & ALL_NUMERIC) or example.exc_msg:
+            # Fallback to base implementation if no NUMERIC flag
+            # set or there was an exception.
+            return doctest.OutputChecker.output_difference(
+                self, example, got, optionflags)
+        else:
+            # Result of base implementation
+            flags = optionflags | NUMERIC_DEFAULT_OPTIONFLAGS
+            base_result = doctest.OutputChecker.output_difference(
+                self, example, got, flags)
+
+            # Our compare result
+            compare_result = self.formatted_compare_numeric(
+                example.want, got, optionflags)
+
+            # Concatenate together
+            return base_result + '\nReason for failure: ' + compare_result + '\n'
+
+
+def run_doctests(verbose=False, print_info=True):
+    failed, attempted = 0, 0
+
+    finder = doctest.DocTestFinder()
+
+    # Use the default docTest.OutputChecker to test our NumericOutputChecker
+    runner = doctest.DocTestRunner(verbose=verbose)
+    for test in finder.find(NumericOutputChecker):
+        runner.run(test)
+    result = runner.summarize()
+    failed += result.failed
+    attempted += result.attempted
+
+    # Test our NumericOutputChecker in action!
+    runner = doctest.DocTestRunner(checker=NumericOutputChecker(), verbose=verbose)
+    for test in finder.find(NumericExample):
+        runner.run(test)
+        result = runner.summarize()
+        failed += result.failed
+        attempted += result.attempted
+
+    return doctest.TestResults(failed, attempted)
+
+run_doctests.__name__ = __name__

snappy/pari.py

@@ -0,0 +1,41 @@
+"""
+Import pari and associated classes and functions here, to be more DRY,
+while supporting both old and new versions of cypari and sage.pari and
+accounting for all of the various idiosyncrasies.
+"""
+
+from packaging.version import parse as parse_version
+from .sage_helper import _within_sage
+
+if _within_sage:
+    from sage.version import version as s_version
+    sage_version = parse_version(s_version)
+    if sage_version < parse_version('9.0'):
+        raise ValueError("you need a more recent version of SageMath")
+    from sage.libs.pari import pari
+    from cypari2 import Gen
+    from cypari2.pari_instance import (prec_words_to_bits,
+                                       prec_bits_to_dec,
+                                       prec_dec_to_bits)
+    try:
+        from sage.all import PariError
+    except ImportError:
+        from sage.libs.pari.all import PariError
+
+    shut_up = lambda: None
+    speak_up = lambda: None
+
+else:  # Plain Python, use CyPari
+    import cypari
+    cypari_version = parse_version(cypari.__version__)
+    if cypari_version < parse_version('2.3'):
+        raise ValueError("you need a more recent version of CyPari")
+    from cypari import pari
+    from cypari._pari import (Gen,
+                              PariError,
+                              prec_words_to_dec,
+                              prec_words_to_bits,
+                              prec_bits_to_dec,
+                              prec_dec_to_bits)
+    shut_up = lambda: pari.shut_up()
+    speak_up = lambda: pari.speak_up()

snappy/phone_home.py

@@ -0,0 +1,57 @@
+import sys
+from threading import Thread
+from .version import version as current
+from packaging.version import parse as parse_version
+
+import ssl
+from urllib import request
+version_url = 'http://snappy.computop.org/current.txt'
+
+
+class Phoner(Thread):
+    def __init__(self):
+        Thread.__init__(self)
+        self.answer = None
+
+    def run(self):
+        this_version = parse_version(current)
+        latest = None
+        try:
+            # Below code would allow switching to https urls in our
+            # macOS app where certs are broken.
+            #
+            # ctx = ssl.create_default_context()
+            # ctx.check_hostname = False
+            # ctx.verify_mode = ssl.CERT_NONE
+            # with request.urlopen(version_url, context=ctx) as response:
+            with request.urlopen(version_url) as response:
+                latest = response.read().decode('ascii').strip()
+                latest_version = parse_version(latest)
+        except Exception:
+            return
+        if latest and latest_version > this_version:
+            self.answer = (latest, current)
+
+
+def update_needed():
+    ET = Phoner()
+    ET.start()
+    ET.join(1.0)  # Wait up to 1.0 seconds for an answer
+    if ET.is_alive():
+        return ''
+    elif ET.answer:
+        return ("**Please upgrade to %s from %s via "
+                "http://snappy.computop.org**\n" % ET.answer)
+    else:
+        return ''
+
+
+if __name__ == '__main__':
+    if len(sys.argv) > 1:
+        current = '0.0.0'
+    else:
+        print('To provoke an update response add an argument.')
+    import time
+    start = time.time()
+    print(update_needed())
+    print('Elapsed time: %ss' % (time.time() - start))