On Fri, Apr 12, 2024 at 7:11 AM Luca Vizzarro <luca.vizza...@arm.com> wrote:
>
<snip>
> @@ -0,0 +1,147 @@
> +# SPDX-License-Identifier: BSD-3-Clause
> +# Copyright(c) 2024 Arm Limited
> +
> +"""Parsing utility module.
> +
> +This module provides :class:`~TextParser` which can be used to model any
> data structure
> +that can parse a block of text.
> +"""
> +
It would be helpful if this top level docstring explained more of how
to use the text parser and some examples of using a small dataclass
that chains some of these methods together. At first glance it wasn't
clear to me why things were done the way they were or what these
methods really provided, but looking at how you use it in the testpmd
shell made more sense.
> +from dataclasses import dataclass, fields, MISSING
> +import re
> +from typing import TypeVar
> +from typing_extensions import Self
> +
> +T = TypeVar("T")
This is declared but I don't see where it is used. There aren't many
typehints in this file, since these types can essentially be anything,
maybe using this could make some things slightly more clear (like, for
example, in eq v1 and v2 are the same type) but in most cases it's
implied so I'm not sure how beneficial this would be regardless.
> +
> +
> +META_PARSERS = "parsers"
> +
> +
> +def chain(parser, metadata):
> + """Chain a parser function.
> +
> + The parser function can take and return a single argument of any type.
> It is
> + up to the user to ensure that the chained functions have compatible
> types.
> +
> + Args:
> + parser: the parser function pointer
> + metadata: pre-existing metadata to chain if any
> + """
> + parsers = metadata.get(META_PARSERS) or []
> + parsers.append(parser)
> + return {**metadata, META_PARSERS: parsers}
> +
> +
> +def to_int(metadata={}, base=0):
Is it simpler to default this to base 10? I assume that's what it'll
be most of the time so we might as well allow users to skip this
parameter.
> + """Converts a string to an integer.
> +
> + Args:
> + metadata: pre-existing metadata to chain if any
> + base: argument passed to the constructor of ``int``
> + """
> + return chain(lambda v: int(v, base), metadata)
> +
<snip>
> +@dataclass
> +class TextParser:
> + """Helper abstract dataclass that parses a text according to the fields'
> rules.
> +
> + This class is accompanied by a selection of parser functions and a
> generic chaining function,
> + that are to be set to the fields' metadata, to enable parsing. If a
> field metadata is not set with
> + any parser function, this is skipped.
> + """
> +
> + @classmethod
> + def parse(cls, text: str) -> Self:
> + """The parsing class method.
> +
> + This function loops through every field that has any parser function
> associated with it and runs
> + each parser chain to the supplied text. If a parser function returns
> None, it expects that parsing
> + has failed and continues to the next field.
> +
> + Args:
> + text: the text to parse
> + Raises:
> + RuntimeError: if the parser did not find a match and the field
> does not have a default value
> + or default factory.
> + """
> + fields_values = {}
> + for field in fields(cls):
> + parsers = field.metadata.get(META_PARSERS)
> + if parsers is None:
> + continue
> +
> + field_value = text
> + for parser_fn in parsers:
> + field_value = parser_fn(field_value)
> + if field_value is None:
> + # nothing was actually parsed, move on
> + break
> +
> + if field_value is None:
> + if field.default is MISSING and field.default_factory is
> MISSING:
> + raise RuntimeError(
> + f"parsers for field {field.name} returned None, but
> the field has no default"
> + )
If we just skip instead of raising an exception here, would this solve
the issues caused by the first and last line in the testpmd output?
The check to see if the first line is an invalid port would obviously
still not work, but would it solve the problem of the trailing prompt?
> + else:
> + fields_values[field.name] = field_value
> +
> + return cls(**fields_values)
> --
> 2.34.1
>
