flexparser 0.3

Parsing made fun ... using typing.

flexparser

Why write another parser? I have asked myself the same question while working on this project. It is clear that there are excellent parsers out there but I wanted to experiment with another way of writing them.

The idea is quite simple. You write a class for every type of content (called here ParsedStatement) you need to parse. Each class should have a from_string constructor. We used extensively the typing module to make the output structure easy to use and less error prone.

For example:

from dataclasses import dataclass

import flexparser as fp

@dataclass(frozen=True)
class Assigment(fp.ParsedStatement):
    """Parses the following `this <- other`
    """

    lhs: str
    rhs: str

    @classmethod
    def from_string(cls, s):
        lhs, rhs = s.split("<-")
        return cls(lhs.strip(), rhs.strip())

(using a frozen dataclass is not necessary but it convenient. Being a dataclass you get the init, str, repr, etc for free. Being frozen, sort of immutable, makes them easier to reason around)

In certain cases you might want to signal the parser that his class is not appropriate to parse the statement.

@dataclass(frozen=True)
class Assigment(fp.ParsedStatement):
    """Parses the following `this <- other`
    """

    lhs: str
    rhs: str

    @classmethod
    def from_string(cls, s):
        if "<-" not in s:
            # This means: I do not know how to parse it
            # try with another ParsedStatement class.
            return None
        lhs, rhs = s.split("<-")
        return cls(lhs.strip(), rhs.strip())

You might also want to indicate that this is the right ParsedStatement but something is not right:

@dataclass(frozen=True)
class InvalidIdentifier(fp.ParsingError):
    value: str


@dataclass(frozen=True)
class Assigment(fp.ParsedStatement):
    """Parses the following `this <- other`
    """

    lhs: str
    rhs: str

    @classmethod
    def from_string(cls, s):
        if "<-" not in s:
            # This means: I do not know how to parse it
            # try with another ParsedStatement class.
            return None
        lhs, rhs = (p.strip() for p in s.split("<-"))

        if not str.isidentifier(lhs):
            return InvalidIdentifier(lhs)

        return cls(lhs, rhs)

Put this into source.txt

one <- other
2two <- new
three <- newvalue
one == three

and then run the following code:

parsed = fp.parse("source.txt", Assigment)
for el in parsed.iter_statements():
    print(repr(el))

will produce the following output:

BOF(start_line=0, start_col=0, end_line=0, end_col=0, raw=None, content_hash=Hash(algorithm_name='blake2b', hexdigest='37bc23cde7cad3ece96b7abf64906c84decc116de1e0486679eb6ca696f233a403f756e2e431063c82abed4f0e342294c2fe71af69111faea3765b78cb90c03f'), path=PosixPath('/Users/grecco/Documents/code/flexparser/examples/in_readme/source1.txt'), mtime=1658550284.9419456)
Assigment(start_line=1, start_col=0, end_line=1, end_col=12, raw='one <- other', lhs='one', rhs='other')
InvalidIdentifier(start_line=2, start_col=0, end_line=2, end_col=11, raw='2two <- new', value='2two')
Assigment(start_line=3, start_col=0, end_line=3, end_col=17, raw='three <- newvalue', lhs='three', rhs='newvalue')
UnknownStatement(start_line=4, start_col=0, end_line=4, end_col=12, raw='one == three')
EOS(start_line=5, start_col=0, end_line=5, end_col=0, raw=None)

The result is a collection of ParsedStatement or ParsingError (flanked by BOF and EOS indicating beginning of file and ending of stream respectively Alternative, it can beginning with BOR with means beginning of resource and it is used when parsing a Python Resource provided with a package).

Notice that there are two correctly parsed statements (Assigment), one error found (InvalidIdentifier) and one unknown (UnknownStatement).

Cool, right? Just writing a from_string method that outputs a datastructure produces a usable structure of parsed objects.

Now what? Let’s say we want to support equality comparison. Simply do:

@dataclass(frozen=True)
class EqualityComparison(fp.ParsedStatement):
    """Parses the following `this == other`
    """

    lhs: str
    rhs: str

    @classmethod
    def from_string(cls, s):
        if "==" not in s:
            return None
        lhs, rhs = (p.strip() for p in s.split("=="))

        return cls(lhs, rhs)

parsed = fp.parse("source.txt", (Assigment, Equality))
for el in parsed.iter_statements():
    print(repr(el))

and run it again:

BOF(start_line=0, start_col=0, end_line=0, end_col=0, raw=None, content_hash=Hash(algorithm_name='blake2b', hexdigest='37bc23cde7cad3ece96b7abf64906c84decc116de1e0486679eb6ca696f233a403f756e2e431063c82abed4f0e342294c2fe71af69111faea3765b78cb90c03f'), path=PosixPath('/Users/grecco/Documents/code/flexparser/examples/in_readme/source1.txt'), mtime=1658550284.9419456)
Assigment(start_line=1, start_col=0, end_line=1, end_col=12, raw='one <- other', lhs='one', rhs='other')
InvalidIdentifier(start_line=2, start_col=0, end_line=2, end_col=11, raw='2two <- new', value='2two')
Assigment(start_line=3, start_col=0, end_line=3, end_col=17, raw='three <- newvalue', lhs='three', rhs='newvalue')
EqualityComparison(start_line=4, start_col=0, end_line=4, end_col=12, raw='one == three', lhs='one', rhs='three')
EOS(start_line=5, start_col=0, end_line=5, end_col=0, raw=None)

You need to group certain statements together: welcome to Block This construct allows you to group

class Begin(fp.ParsedStatement):

    @classmethod
    def from_string(cls, s):
        if s == "begin":
            return cls()

        return None

class End(fp.ParsedStatement):

    @classmethod
    def from_string(cls, s):
        if s == "end":
            return cls()

        return None

class AssigmentBlock(fp.Block[Begin, Assigment, End]):
    pass

parsed = fp.parse("source.txt", (AssigmentBlock, Equality))

Run the code:

BOF(start_line=0, start_col=0, end_line=0, end_col=0, raw=None, content_hash=Hash(algorithm_name='blake2b', hexdigest='37bc23cde7cad3ece96b7abf64906c84decc116de1e0486679eb6ca696f233a403f756e2e431063c82abed4f0e342294c2fe71af69111faea3765b78cb90c03f'), path=PosixPath('/Users/grecco/Documents/code/flexparser/examples/in_readme/source1.txt'), mtime=1658550284.9419456)
UnknownStatement(start_line=1, start_col=0, end_line=1, end_col=12, raw='one <- other')
UnknownStatement(start_line=2, start_col=0, end_line=2, end_col=11, raw='2two <- new')
UnknownStatement(start_line=3, start_col=0, end_line=3, end_col=17, raw='three <- newvalue')
UnknownStatement(start_line=4, start_col=0, end_line=4, end_col=12, raw='one == three')
EOS(start_line=5, start_col=0, end_line=5, end_col=0, raw=None)

Notice that there are a lot of UnknownStatement now, because we instructed the parser to only look for assignment within a block. So change your text file to:

begin
one <- other
2two <- new
three <- newvalue
end
one == three

and try again:

BOF(start_line=0, start_col=0, end_line=0, end_col=0, raw=None, content_hash=Hash(algorithm_name='blake2b', hexdigest='3d8ce0051dcdd6f0f80ef789a0df179509d927874f242005ac41ed886ae0b71a30b845b9bfcb30194461c0ef6a3ca324c36f411dfafc7e588611f1eb0269bb5a'), path=PosixPath('/Users/grecco/Documents/code/flexparser/examples/in_readme/source2.txt'), mtime=1658550707.1248093)
Begin(start_line=1, start_col=0, end_line=1, end_col=5, raw='begin')
Assigment(start_line=2, start_col=0, end_line=2, end_col=12, raw='one <- other', lhs='one', rhs='other')
InvalidIdentifier(start_line=3, start_col=0, end_line=3, end_col=11, raw='2two <- new', value='2two')
Assigment(start_line=4, start_col=0, end_line=4, end_col=17, raw='three <- newvalue', lhs='three', rhs='newvalue')
End(start_line=5, start_col=0, end_line=5, end_col=3, raw='end')
EqualityComparison(start_line=6, start_col=0, end_line=6, end_col=12, raw='one == three', lhs='one', rhs='three')
EOS(start_line=7, start_col=0, end_line=7, end_col=0, raw=None)

Until now we have used parsed.iter_statements to iterate over all parsed statements. But let’s look inside parsed, an object of ParsedProject type. It is a thin wrapper over a dictionary mapping files to parsed content. Because we have provided a single file and this does not contain a link another, our parsed object contains a single element. The key is None indicating that the file ‘source.txt’ was loaded from the root location (None). The content is a ParsedSourceFile object with the following attributes:

• path: full path of the source file

• mtime: modification file of the source file

• content_hash: hash of the pickled content

• config: extra parameters that can be given to the parser (see below).

ParsedSource(
    parsed_source=parse.<locals>.CustomRootBlock(
        opening=BOF(start_line=0, start_col=0, end_line=0, end_col=0, raw=None, content_hash=Hash(algorithm_name='blake2b', hexdigest='3d8ce0051dcdd6f0f80ef789a0df179509d927874f242005ac41ed886ae0b71a30b845b9bfcb30194461c0ef6a3ca324c36f411dfafc7e588611f1eb0269bb5a'), path=PosixPath('/Users/grecco/Documents/code/flexparser/examples/in_readme/source2.txt'), mtime=1658550707.1248093),
        body=(
            Block.subclass_with.<locals>.CustomBlock(
                opening=Begin(start_line=1, start_col=0, end_line=1, end_col=5, raw='begin'),
                body=(
                    Assigment(start_line=2, start_col=0, end_line=2, end_col=12, raw='one <- other', lhs='one', rhs='other'),
                    InvalidIdentifier(start_line=3, start_col=0, end_line=3, end_col=11, raw='2two <- new', value='2two'),
                    Assigment(start_line=4, start_col=0, end_line=4, end_col=17, raw='three <- newvalue', lhs='three', rhs='newvalue')
                ),
                closing=End(start_line=5, start_col=0, end_line=5, end_col=3, raw='end')),
            EqualityComparison(start_line=6, start_col=0, end_line=6, end_col=12, raw='one == three', lhs='one', rhs='three')),
        closing=EOS(start_line=7, start_col=0, end_line=7, end_col=0, raw=None)),
    config=None
)

A few things to notice:

1. We were using a block before without knowing. The RootBlock is a special type of Block that starts and ends automatically with the file.

2. opening, body, closing are automatically annotated with the possible ParsedStatement (plus ParsingError), therefore autocompletes works in most IDEs.

3. The same is true for the defined ParsedStatement (we have use dataclass for a reason). This makes using the actual result of the parsing a charm!.

4. That annoying subclass_with.<locals> is because we have built a class on the fly when we used Block.subclass_with. You can get rid of it (which is actually useful for pickling) by explicit subclassing Block in your code (see below).

Multiple source files

Most projects have more than one source file internally connected. A file might refer to another that also need to be parsed (e.g. an #include statement in c). flexparser provides the IncludeStatement base class specially for this purpose.

@dataclass(frozen=True)
class Include(fp.IncludeStatement):
    """A naive implementation of #include "file"
    """

    value: str

    @classmethod
    def from_string(cls, s):
        if s.startwith("#include "):
            return None

        value = s[len("#include "):].strip().strip('"')

        return cls(value)

    @propery
    def target(self):
        return self.value

The only difference is that you need to implement a target property that returns the file name or resource that this statement refers to.

Customizing statementization

statementi … what? flexparser works by trying to parse each statement with one of the known classes. So it is fair to ask what is an statement in this context and how can you configure it to your needs. A text file is split into non overlapping strings called statements. Parsing work as follows:

1. each file is split into statements (can be single or multi line).

2. each statement is parsed with the first of the contextually available ParsedStatement or Block subclassed that returns a ParsedStatement or ParsingError

You can customize how to split each line into statements with two arguments provided to parse:

• strip_spaces (bool): indicates that leading and trailing spaces must be removed before attempting to parse. (default: True)

• delimiters (dict): indicates how each line must be subsplit. (default: do not divide)

An delimiter example might be {";": (fp.DelimiterInclude.SKIP, fp.DelimiterAction.CONTINUE)} which tells the statementizer (sorry) that when a “;” is found a new statement should begin. DelimiterMode.SKIP tells that “;” should not be added to the previous statement nor to the next. Other valid values are SPLIT_AFTER and SPLIT_BEFORE to append or prepend the delimiter character to the previous or next statement. The second element tells the statementizer (sorry again) what to do next: valid values are: CONTINUE, CAPTURE_NEXT_TIL_EOL, STOP_PARSING_LINE, and STOP_PARSING.

This is useful with comments. For example, {"#": (fp.DelimiterMode.WITH_NEXT, fp.DelimiterAction.CAPTURE_NEXT_TIL_EOL))} tells the statementizer (it is not funny anymore) that after the first “#” it should stop splitting and capture all.

This allows:

## This will work as a single statement
# This will work as a single statement #
# This will work as # a single statement #
a = 3 # this will produce two statements (a=3, and the rest)

Explicit Block classes

class AssigmentBlock(fp.Block[Begin, Assigment, End]):
    pass

class EntryBlock(fp.RootBlock[Union[AssigmentBlock, Equality]]):
    pass

parsed = fp.parse("source.txt", EntryBlock)

Customizing parsing

In certain cases you might want to leave to the user some configuration details. We have method for that!. Instead of overriding from_string override from_string_and_config. The second argument is an object that can be given to the parser, which in turn will be passed to each ParsedStatement class.

@dataclass(frozen=True)
class NumericAssigment(fp.ParsedStatement):
    """Parses the following `this <- other`
    """

    lhs: str
    rhs: numbers.Number

    @classmethod
    def from_string_and_config(cls, s, config):
        if "==" not in s:
            # This means: I do not know how to parse it
            # try with another ParsedStatement class.
            return None
        lhs, rhs = s.split("==")
        return cls(lhs.strip(), config.numeric_type(rhs.strip()))

class Config:

    numeric_type = float

parsed = fp.parse("source.txt", NumericAssigment, Config)

---

This project was started as a part of Pint, the python units package.

See AUTHORS for a list of the maintainers.

To review an ordered list of notable changes for each version of a project, see CHANGES