import math
import textwrap
from typing import Optional

import torch
from torch import inf


class __PrinterOptions:
    precision: int = 4
    threshold: float = 1000
    edgeitems: int = 3
    linewidth: int = 80
    sci_mode: Optional[bool] = None


PRINT_OPTS = __PrinterOptions()


# We could use **kwargs, but this will give better docs
def set_printoptions(
    precision=None,
    threshold=None,
    edgeitems=None,
    linewidth=None,
    profile=None,
    sci_mode=None,
):
    r"""Set options for printing. Items shamelessly taken from NumPy

    Args:
        precision: Number of digits of precision for floating point output
            (default = 4).
        threshold: Total number of array elements which trigger summarization
            rather than full `repr` (default = 1000).
        edgeitems: Number of array items in summary at beginning and end of
            each dimension (default = 3).
        linewidth: The number of characters per line for the purpose of
            inserting line breaks (default = 80). Thresholded matrices will
            ignore this parameter.
        profile: Sane defaults for pretty printing. Can override with any of
            the above options. (any one of `default`, `short`, `full`)
        sci_mode: Enable (True) or disable (False) scientific notation. If
            None (default) is specified, the value is defined by
            `torch._tensor_str._Formatter`. This value is automatically chosen
            by the framework.

    Example::

        >>> # Limit the precision of elements
        >>> torch.set_printoptions(precision=2)
        >>> torch.tensor([1.12345])
        tensor([1.12])
        >>> # Limit the number of elements shown
        >>> torch.set_printoptions(threshold=5)
        >>> torch.arange(10)
        tensor([0, 1, 2, ..., 7, 8, 9])
        >>> # Restore defaults
        >>> torch.set_printoptions(profile='default')
        >>> torch.tensor([1.12345])
        tensor([1.1235])
        >>> torch.arange(10)
        tensor([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])

    """
    if profile is not None:
        if profile == "default":
            PRINT_OPTS.precision = 4
            PRINT_OPTS.threshold = 1000
            PRINT_OPTS.edgeitems = 3
            PRINT_OPTS.linewidth = 80
        elif profile == "short":
            PRINT_OPTS.precision = 2
            PRINT_OPTS.threshold = 1000
            PRINT_OPTS.edgeitems = 2
            PRINT_OPTS.linewidth = 80
        elif profile == "full":
            PRINT_OPTS.precision = 4
            PRINT_OPTS.threshold = inf
            PRINT_OPTS.edgeitems = 3
            PRINT_OPTS.linewidth = 80

    if precision is not None:
        PRINT_OPTS.precision = precision
    if threshold is not None:
        PRINT_OPTS.threshold = threshold
    if edgeitems is not None:
        PRINT_OPTS.edgeitems = edgeitems
    if linewidth is not None:
        PRINT_OPTS.linewidth = linewidth
    PRINT_OPTS.sci_mode = sci_mode


def tensor_totype(t):
    dtype = torch.float if t.is_mps else torch.double
    return t.to(dtype=dtype)


class _Formatter:
    def __init__(self, tensor):
        self.floating_dtype = tensor.dtype.is_floating_point
        self.int_mode = True
        self.sci_mode = False
        self.max_width = 1

        with torch.no_grad():
            tensor_view = tensor.reshape(-1)

        if not self.floating_dtype:
            for value in tensor_view:
                value_str = "{}".format(value)
                self.max_width = max(self.max_width, len(value_str))

        else:
            nonzero_finite_vals = torch.masked_select(
                tensor_view, torch.isfinite(tensor_view) & tensor_view.ne(0)
            )

            if nonzero_finite_vals.numel() == 0:
                # no valid number, do nothing
                return

            # Convert to double for easy calculation. HalfTensor overflows with 1e8, and there's no div() on CPU.
            nonzero_finite_abs = tensor_totype(nonzero_finite_vals.abs())
            nonzero_finite_min = tensor_totype(nonzero_finite_abs.min())
            nonzero_finite_max = tensor_totype(nonzero_finite_abs.max())

            for value in nonzero_finite_vals:
                if value != torch.ceil(value):
                    self.int_mode = False
                    break

            if self.int_mode:
                # in int_mode for floats, all numbers are integers, and we append a decimal to nonfinites
                # to indicate that the tensor is of floating type. add 1 to the len to account for this.
                if (
                    nonzero_finite_max / nonzero_finite_min > 1000.0
                    or nonzero_finite_max > 1.0e8
                ):
                    self.sci_mode = True
                    for value in nonzero_finite_vals:
                        value_str = (
                            ("{{:.{}e}}").format(PRINT_OPTS.precision).format(value)
                        )
                        self.max_width = max(self.max_width, len(value_str))
                else:
                    for value in nonzero_finite_vals:
                        value_str = ("{:.0f}").format(value)
                        self.max_width = max(self.max_width, len(value_str) + 1)
            else:
                # Check if scientific representation should be used.
                if (
                    nonzero_finite_max / nonzero_finite_min > 1000.0
                    or nonzero_finite_max > 1.0e8
                    or nonzero_finite_min < 1.0e-4
                ):
                    self.sci_mode = True
                    for value in nonzero_finite_vals:
                        value_str = (
                            ("{{:.{}e}}").format(PRINT_OPTS.precision).format(value)
                        )
                        self.max_width = max(self.max_width, len(value_str))
                else:
                    for value in nonzero_finite_vals:
                        value_str = (
                            ("{{:.{}f}}").format(PRINT_OPTS.precision).format(value)
                        )
                        self.max_width = max(self.max_width, len(value_str))

        if PRINT_OPTS.sci_mode is not None:
            self.sci_mode = PRINT_OPTS.sci_mode

    def width(self):
        return self.max_width

    def format(self, value):
        if self.floating_dtype:
            if self.sci_mode:
                ret = (
                    ("{{:{}.{}e}}")
                    .format(self.max_width, PRINT_OPTS.precision)
                    .format(value)
                )
            elif self.int_mode:
                ret = "{:.0f}".format(value)
                if not (math.isinf(value) or math.isnan(value)):
                    ret += "."
            else:
                ret = ("{{:.{}f}}").format(PRINT_OPTS.precision).format(value)
        else:
            ret = "{}".format(value)
        return (self.max_width - len(ret)) * " " + ret


def _scalar_str(self, formatter1, formatter2=None):
    if formatter2 is not None:
        real_str = _scalar_str(self.real, formatter1)
        imag_str = (_scalar_str(self.imag, formatter2) + "j").lstrip()
        # handles negative numbers, +0.0, -0.0
        if imag_str[0] == "+" or imag_str[0] == "-":
            return real_str + imag_str
        else:
            return real_str + "+" + imag_str
    else:
        return formatter1.format(self.item())


def _vector_str(self, indent, summarize, formatter1, formatter2=None):
    # length includes spaces and comma between elements
    element_length = formatter1.width() + 2
    if formatter2 is not None:
        # width for imag_formatter + an extra j for complex
        element_length += formatter2.width() + 1

    elements_per_line = max(
        1, int(math.floor((PRINT_OPTS.linewidth - indent) / (element_length)))
    )

    def _val_formatter(val, formatter1=formatter1, formatter2=formatter2):
        if formatter2 is not None:
            real_str = formatter1.format(val.real)
            imag_str = (formatter2.format(val.imag) + "j").lstrip()
            # handles negative numbers, +0.0, -0.0
            if imag_str[0] == "+" or imag_str[0] == "-":
                return real_str + imag_str
            else:
                return real_str + "+" + imag_str
        else:
            return formatter1.format(val)

    if summarize and self.size(0) > 2 * PRINT_OPTS.edgeitems:
        data = (
            [_val_formatter(val) for val in self[: PRINT_OPTS.edgeitems].tolist()]
            + [" ..."]
            + [_val_formatter(val) for val in self[-PRINT_OPTS.edgeitems :].tolist()]
        )
    else:
        data = [_val_formatter(val) for val in self.tolist()]

    data_lines = [
        data[i : i + elements_per_line] for i in range(0, len(data), elements_per_line)
    ]
    lines = [", ".join(line) for line in data_lines]
    return "[" + ("," + "\n" + " " * (indent + 1)).join(lines) + "]"


# formatter2 is only used for printing complex tensors.
# For complex tensors, formatter1 and formatter2 are the formatters for tensor.real
# and tensor.imag respesectively
def _tensor_str_with_formatter(self, indent, summarize, formatter1, formatter2=None):
    dim = self.dim()

    if dim == 0:
        return _scalar_str(self, formatter1, formatter2)

    if dim == 1:
        return _vector_str(self, indent, summarize, formatter1, formatter2)

    if summarize and self.size(0) > 2 * PRINT_OPTS.edgeitems:
        slices = (
            [
                _tensor_str_with_formatter(
                    self[i], indent + 1, summarize, formatter1, formatter2
                )
                for i in range(0, PRINT_OPTS.edgeitems)
            ]
            + ["..."]
            + [
                _tensor_str_with_formatter(
                    self[i], indent + 1, summarize, formatter1, formatter2
                )
                for i in range(len(self) - PRINT_OPTS.edgeitems, len(self))
            ]
        )
    else:
        slices = [
            _tensor_str_with_formatter(
                self[i], indent + 1, summarize, formatter1, formatter2
            )
            for i in range(0, self.size(0))
        ]

    tensor_str = ("," + "\n" * (dim - 1) + " " * (indent + 1)).join(slices)
    return "[" + tensor_str + "]"


def _tensor_str(self, indent):
    if self.numel() == 0:
        return "[]"

    if self.has_names():
        # There are two main codepaths (possibly more) that tensor printing goes through:
        # - tensor data can fit comfortably on screen
        # - tensor data needs to be summarized
        # Some of the codepaths don't fully support named tensors, so we send in
        # an unnamed tensor to the formatting code as a workaround.
        self = self.rename(None)

    summarize = self.numel() > PRINT_OPTS.threshold

    if self._is_zerotensor():
        self = self.clone()

    # handle the negative bit
    if self.is_neg():
        self = self.resolve_neg()

    if self.dtype is torch.float16 or self.dtype is torch.bfloat16:
        self = self.float()

    if self.dtype is torch.complex32:
        self = self.cfloat()

    if self.dtype.is_complex:
        # handle the conjugate bit
        self = self.resolve_conj()
        real_formatter = _Formatter(
            get_summarized_data(self.real) if summarize else self.real
        )
        imag_formatter = _Formatter(
            get_summarized_data(self.imag) if summarize else self.imag
        )
        return _tensor_str_with_formatter(
            self, indent, summarize, real_formatter, imag_formatter
        )
    else:
        formatter = _Formatter(get_summarized_data(self) if summarize else self)
        return _tensor_str_with_formatter(self, indent, summarize, formatter)


def _add_suffixes(tensor_str, suffixes, indent, force_newline):
    tensor_strs = [tensor_str]
    last_line_len = len(tensor_str) - tensor_str.rfind("\n") + 1
    for suffix in suffixes:
        suffix_len = len(suffix)
        if force_newline or last_line_len + suffix_len + 2 > PRINT_OPTS.linewidth:
            tensor_strs.append(",\n" + " " * indent + suffix)
            last_line_len = indent + suffix_len
            force_newline = False
        else:
            tensor_strs.append(", " + suffix)
            last_line_len += suffix_len + 2
    tensor_strs.append(")")
    return "".join(tensor_strs)