我也需要那个，对原来的不满意pprint。具体来说，我希望它进行正常缩进（2 或 4 个空格），而不是缩进pprint。
具体来说，对于某些 dict，我得到了原始输出pprint：
{'melgan': {'class': 'subnetwork',
'from': 'data',
'subnetwork': {'l0': {'axes': 'spatial',
'class': 'pad',
'from': 'data',
'mode': 'reflect',
'padding': (3, 3)},
'la1': {'activation': None,
'class': 'conv',
'dilation_rate': (1,),
'filter_size': (7,),
'from': 'l0',
'n_out': 384,
'padding': 'valid',
'strides': (1,),
'with_bias': True},
'lay2': {'class': 'eval',
'eval': 'tf.nn.leaky_relu(source(0), '
'alpha=0.2)',
'from': 'la1'},
'layer3_xxx': {'activation': None,
'class': 'transposed_conv',
'filter_size': (10,),
'from': 'lay2',
'n_out': 192,
'output_padding': (1,),
'padding': 'valid',
'remove_padding': (3,),
'strides': (5,),
'with_bias': True},
'output': {'class': 'copy', 'from': 'layer3_xxx'}}},
'output': {'class': 'copy', 'from': 'melgan'}}
但我希望它是这样的：
{
'melgan': {
'class': 'subnetwork',
'from': 'data',
'subnetwork': {
'l0': {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (3, 3), 'from': 'data'},
'la1': {
'class': 'conv',
'from': 'l0',
'activation': None,
'with_bias': True,
'n_out': 384,
'filter_size': (7,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'lay2': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'la1'},
'layer3_xxx': {
'class': 'transposed_conv',
'from': 'lay2',
'activation': None,
'with_bias': True,
'n_out': 192,
'filter_size': (10,),
'strides': (5,),
'padding': 'valid',
'output_padding': (1,),
'remove_padding': (3,)
},
'output': {'class': 'copy', 'from': 'layer3_xxx'}
}
},
'output': {'class': 'copy', 'from': 'melgan'}
}
有Python Rich library，它还提供了一个自己的pprint变体 ( from rich.pretty import pprint)，这与我想要的很接近。
还有pprintpp，也很接近。
我在这里实现了一个非常简单的变体。代码：
from typing import Any
import sys
import numpy
def pprint(o: Any, *, file=sys.stdout,
prefix="", postfix="",
line_prefix="", line_postfix="\n") -> None:
if "\n" in line_postfix and _type_simplicity_score(o) <= _type_simplicity_limit:
prefix = f"{line_prefix}{prefix}"
line_prefix = ""
postfix = postfix + line_postfix
line_postfix = ""
def _sub_pprint(o: Any, prefix="", postfix="", inc_indent=True):
multi_line = "\n" in line_postfix
if not multi_line and postfix.endswith(","):
postfix += " "
pprint(
o, file=file, prefix=prefix, postfix=postfix,
line_prefix=(line_prefix + " " * inc_indent) if multi_line else "",
line_postfix=line_postfix)
def _print(s: str, is_end: bool = False):
nonlocal prefix # no need for is_begin, just reset prefix
file.write(line_prefix)
file.write(prefix)
file.write(s)
if is_end:
file.write(postfix)
file.write(line_postfix)
if "\n" in line_postfix:
file.flush()
prefix = ""
def _print_list():
for i, v in enumerate(o):
_sub_pprint(v, postfix="," if i < len(o) - 1 else "")
if isinstance(o, list):
if len(o) == 0:
_print("[]", is_end=True)
return
_print("[")
_print_list()
_print("]", is_end=True)
return
if isinstance(o, tuple):
if len(o) == 0:
_print("()", is_end=True)
return
if len(o) == 1:
_sub_pprint(o[0], prefix=f"{prefix}(", postfix=f",){postfix}", inc_indent=False)
return
_print("(")
_print_list()
_print(")", is_end=True)
return
if isinstance(o, set):
if len(o) == 0:
_print("set()", is_end=True)
return
_print("{")
_print_list()
_print("}", is_end=True)
return
if isinstance(o, dict):
if len(o) == 0:
_print("{}", is_end=True)
return
_print("{")
for i, (k, v) in enumerate(o.items()):
_sub_pprint(v, prefix=f"{k!r}: ", postfix="," if i < len(o) - 1 else "")
_print("}", is_end=True)
return
if isinstance(o, numpy.ndarray):
_sub_pprint(
o.tolist(),
prefix=f"{prefix}numpy.array(",
postfix=f", dtype=numpy.{o.dtype}){postfix}",
inc_indent=False)
return
# fallback
_print(repr(o), is_end=True)
def pformat(o: Any) -> str:
import io
s = io.StringIO()
pprint(o, file=s)
return s.getvalue()
_type_simplicity_limit = 120. # magic number
def _type_simplicity_score(o: Any, _offset=0.) -> float:
"""
:param Any o:
:param float _offset:
:return: a score, which is a very rough estimate of len(repr(o)), calculated efficiently
"""
_spacing = 2.
if isinstance(o, bool):
return 4. + _offset
if isinstance(o, (int, numpy.integer)):
if o == 0:
return 1. + _offset
return 1. + numpy.log10(abs(o)) + _offset
if isinstance(o, str):
return 2. + len(o) + _offset
if isinstance(o, (float, complex, numpy.number)):
return len(repr(o)) + _offset
if isinstance(o, (tuple, list, set)):
for x in o:
_offset = _type_simplicity_score(x, _offset=_offset + _spacing)
if _offset > _type_simplicity_limit:
break
return _offset
if isinstance(o, dict):
for x in o.values(): # ignore keys...
_offset = _type_simplicity_score(x, _offset=_offset + 10. + _spacing) # +10 for key
if _offset > _type_simplicity_limit:
break
return _offset
if isinstance(o, numpy.ndarray):
_offset += 10. # prefix/postfix
if o.size * 2. + _offset > _type_simplicity_limit: # too big already?
return o.size * 2. + _offset
if str(o.dtype).startswith("int"):
a = _type_simplicity_score(numpy.max(numpy.abs(o))) + _spacing
return o.size * a + _offset
a = max([_type_simplicity_score(x) for x in o.flatten()]) + _spacing
return o.size * a + _offset
# Unknown object. Fallback > _type_simplicity_limit.
return _type_simplicity_limit + 1. + _offset