builders and pipeline

5 files changed, 368 insertions, 1 deletions

diff --git a/bsie/base/errors.py b/bsie/base/errors.py
index a86b7e8..760351f 100644
--- a/bsie/base/errors.py
+++ b/bsie/base/errors.py
@@ -9,7 +9,9 @@ import typing
 
 # exports
 __all__: typing.Sequence[str] = (
+    'BuilderError',
     'ExtractorError',
+    'LoaderError',
     'ReaderError',
     )
 
@@ -19,6 +21,12 @@ __all__: typing.Sequence[str] = (
 class _BSIEError(Exception):
     """Generic BSIE error."""
 
+class BuilderError(_BSIEError):
+    """The Builder failed to create an instance."""
+
+class LoaderError(BuilderError):
+    """Failed to load a module or class."""
+
 class ExtractorError(_BSIEError):
     """The Extractor failed to process the given content."""
 
diff --git a/bsie/tools/__init__.py b/bsie/tools/__init__.py
new file mode 100644
index 0000000..8ca9620
--- /dev/null
+++ b/bsie/tools/__init__.py
@@ -0,0 +1,20 @@
+"""
+
+Part of the bsie module.
+A copy of the license is provided with the project.
+Author: Matthias Baumgartner, 2022
+"""
+# imports
+import typing
+
+# inner-module imports
+from . import builder
+from . import pipeline
+
+# exports
+__all__: typing.Sequence[str] = (
+    'builder',
+    'pipeline',
+    )
+
+## EOF ##
diff --git a/bsie/tools/builder.py b/bsie/tools/builder.py
new file mode 100644
index 0000000..8f7a410
--- /dev/null
+++ b/bsie/tools/builder.py
@@ -0,0 +1,217 @@
+"""
+
+Part of the bsie module.
+A copy of the license is provided with the project.
+Author: Matthias Baumgartner, 2022
+"""
+# imports
+import importlib
+import logging
+import typing
+
+# bsie imports
+from bsie import base
+from bsie.base import errors
+from bsie.utils.bsfs import URI, typename
+
+# inner-module imports
+from . import pipeline
+
+# exports
+__all__: typing.Sequence[str] = (
+    'ExtractorBuilder',
+    'PipelineBuilder',
+    'ReaderBuilder',
+    )
+
+
+## code ##
+
+logger = logging.getLogger(__name__)
+
+def _safe_load(module_name: str, class_name: str):
+    """Get a class from a module. Raise BuilderError if anything goes wrong."""
+    try:
+        # load the module
+        module = importlib.import_module(module_name)
+    except Exception as err:
+        # cannot import module
+        raise errors.LoaderError(f'cannot load module {module_name}') from err
+
+    try:
+        # get the class from the module
+        cls = getattr(module, class_name)
+    except Exception as err:
+        # cannot find the class
+        raise errors.LoaderError(f'cannot load class {class_name} from module {module_name}') from err
+
+    return cls
+
+
+def _unpack_name(name):
+    """Split a name into its module and class component (dot-separated)."""
+    if not isinstance(name, str):
+        raise TypeError(name)
+    if '.' not in name:
+        raise ValueError('name must be a qualified class name.')
+    module_name, class_name = name[:name.rfind('.')], name[name.rfind('.')+1:]
+    if module_name == '':
+        raise ValueError('name must be a qualified class name.')
+    return module_name, class_name
+
+
+class ReaderBuilder():
+    """Build `bsie.base.reader.Reader` instances.
+
+    Readers are defined via their qualified class name
+    (e.g., bsie.reader.path.Path) and optional keyword
+    arguments that are passed to the constructor via
+    the *kwargs* argument (name as key, kwargs as value).
+    The ReaderBuilder keeps a cache of previously built
+    reader instances, as they are anyway built with
+    identical keyword arguments.
+
+    """
+
+    # keyword arguments
+    kwargs: typing.Dict[str, typing.Dict[str, typing.Any]]
+
+    # cached readers
+    cache: typing.Dict[str, base.reader.Reader]
+
+    def __init__(self, kwargs: typing.Dict[str, typing.Dict[str, typing.Any]]):
+        self.kwargs = kwargs
+        self.cache = {}
+
+    def build(self, name: str) -> base.reader.Reader:
+        """Return an instance for the qualified class name."""
+        # return cached instance
+        if name in self.cache:
+            return self.cache[name]
+
+        # check name and get module/class components
+        module_name, class_name = _unpack_name(name)
+
+        # import reader class
+        cls = _safe_load(module_name, class_name)
+
+        # get kwargs
+        kwargs = self.kwargs.get(name, {})
+        if not isinstance(kwargs, dict):
+            raise TypeError(f'expected a kwargs dict, found {typename(kwargs)}')
+
+        try: # build, cache, and return instance
+            obj = cls(**kwargs)
+            # cache instance
+            self.cache[name] = obj
+            # return instance
+            return obj
+
+        except Exception as err:
+            raise errors.BuilderError(f'failed to build reader {name} due to {typename(err)}: {err}') from err
+
+
+class ExtractorBuilder():
+    """Build `bsie.base.extractor.Extractor instances.
+
+    It is permissible to build multiple instances of the same extractor
+    (typically with different arguments), hence the ExtractorBuilder
+    receives a list of build specifications. Each specification is
+    a dict with a single key (extractor's qualified name) and a dict
+    to be used as keyword arguments.
+    Example: [{'bsie.extractor.generic.path.Path': {}}, ]
+
+    """
+
+    # build specifications
+    specs: typing.List[typing.Dict[str, typing.Dict[str, typing.Any]]]
+
+    def __init__(self, specs: typing.List[typing.Dict[str, typing.Dict[str, typing.Any]]]):
+        self.specs = specs
+
+    def __iter__(self) -> typing.Iterator[int]:
+        """Iterate over extractor specifications."""
+        return iter(range(len(self.specs)))
+
+    def build(self, index: int) -> base.extractor.Extractor:
+        """Return an instance of the n'th extractor (n=*index*)."""
+        # get build instructions
+        specs = self.specs[index]
+
+        # check specs structure. expecting[{name: {kwargs}}]
+        if not isinstance(specs, dict):
+            raise TypeError(f'expected a dict, found {typename(specs)}')
+        if len(specs) != 1:
+            raise TypeError(f'expected a dict of length one, found {len(specs)}')
+
+        # get name and args from specs
+        name = next(iter(specs.keys()))
+        kwargs = specs[name]
+
+        # check kwargs structure
+        if not isinstance(kwargs, dict):
+            raise TypeError(f'expected a dict, found {typename(kwargs)}')
+
+        # check name and get module/class components
+        module_name, class_name = _unpack_name(name)
+
+        # import extractor class
+        cls = _safe_load(module_name, class_name)
+
+        try: # build and return instance
+            return cls(**kwargs)
+
+        except Exception as err:
+            raise errors.BuilderError(f'failed to build extractor {name} due to {typename(err)}: {err}') from err
+
+
+class PipelineBuilder():
+    """Build `bsie.tools.pipeline.Pipeline` instances."""
+
+    def __init__(
+            self,
+            prefix: URI,
+            reader_builder: ReaderBuilder,
+            extractor_builder: ExtractorBuilder,
+            ):
+        self.prefix = prefix
+        self.rbuild = reader_builder
+        self.ebuild = extractor_builder
+
+    def build(self) -> pipeline.Pipeline:
+        """Return a Pipeline instance."""
+        ext2rdr = {}
+
+        for eidx in self.ebuild:
+            # build extractor
+            try:
+                ext = self.ebuild.build(eidx)
+
+            except errors.LoaderError as err: # failed to load extractor; skip
+                logger.error('failed to load extractor: %s', err)
+                continue
+
+            except errors.BuilderError as err: # failed to build instance; skip
+                logger.error(str(err))
+                continue
+
+            try:
+                # get reader required by extractor
+                if ext.CONTENT_READER is not None:
+                    rdr = self.rbuild.build(ext.CONTENT_READER)
+                else:
+                    rdr = None
+                # store extractor
+                ext2rdr[ext] = rdr
+
+            except errors.LoaderError as err: # failed to load reader
+                logger.error('failed to load reader: %s', err)
+
+            except errors.BuilderError as err: # failed to build reader
+                logger.error(str(err))
+
+        return pipeline.Pipeline(self.prefix, ext2rdr)
+
+
+
+## EOF ##
diff --git a/bsie/tools/pipeline.py b/bsie/tools/pipeline.py
new file mode 100644
index 0000000..8e1c992
--- /dev/null
+++ b/bsie/tools/pipeline.py
@@ -0,0 +1,121 @@
+"""
+
+Part of the bsie module.
+A copy of the license is provided with the project.
+Author: Matthias Baumgartner, 2022
+"""
+# imports
+from collections import defaultdict
+import logging
+import typing
+
+# bsie imports
+from bsie import base
+from bsie.utils import ns
+from bsie.utils.node import Node
+from bsie.utils.bsfs import schema as _schema, URI, uuid as _uuid, typename
+
+# exports
+__all__: typing.Sequence[str] = (
+    'Pipeline',
+    )
+
+## code ##
+
+logger = logging.getLogger(__name__)
+
+class Pipeline():
+    """Extraction pipeline to generate triples from files.
+
+    The Pipeline binds readers and extractors, and performs
+    the necessary operations to produce triples from a file.
+    It takes a best-effort approach to extract as many triples
+    as possible. Errors during the extraction are passed over
+    and reported to the log.
+
+    """
+
+    # combined extractor schemas.
+    schema: _schema.Schema
+
+    # node prefix.
+    _prefix: URI
+
+    # extractor -> reader mapping
+    _ext2rdr: typing.Dict[base.extractor.Extractor, typing.Optional[base.reader.Reader]]
+
+    def __init__(
+            self,
+            prefix: URI,
+            ext2rdr: typing.Dict[base.extractor.Extractor, typing.Optional[base.reader.Reader]]
+            ):
+        # store core members
+        self._prefix = prefix
+        self._ext2rdr = ext2rdr
+        # compile schema from all extractors
+        self.schema = _schema.Schema.Union(ext.schema for ext in ext2rdr)
+
+    def __str__(self) -> str:
+        return typename(self)
+
+    def __repr__(self) -> str:
+        return f'{typename(self)}(...)'
+
+    def __hash__(self) -> int:
+        return hash((type(self), self._prefix, self.schema, tuple(self._ext2rdr), tuple(self._ext2rdr.values())))
+
+    def __eq__(self, other: typing.Any) -> bool:
+        return isinstance(other, type(self)) \
+           and self.schema == other.schema \
+           and self._prefix == other._prefix \
+           and self._ext2rdr == other._ext2rdr
+
+    def __call__(
+            self,
+            path: URI,
+            predicates: typing.Optional[typing.Iterable[_schema.Predicate]] = None,
+            ) -> typing.Iterator[typing.Tuple[Node, _schema.Predicate, typing.Any]]:
+        """Extract triples from the file at *path*. Optionally, limit triples to *predicates*."""
+        # get predicates
+        predicates = set(predicates) if predicates is not None else set(self.schema.predicates())
+
+        # get extractors
+        extractors = {ext for ext in self._ext2rdr if not set(ext.predicates()).isdisjoint(predicates)}
+
+        # corner-case short-cut
+        if len(extractors) == 0:
+            return
+
+        # get readers -> extractors mapping
+        rdr2ext = defaultdict(set)
+        for ext in extractors:
+            rdr = self._ext2rdr[ext]
+            rdr2ext[rdr].add(ext)
+
+        # create subject for file
+        uuid = _uuid.UCID.from_path(path)
+        subject = Node(ns.bsfs.Entity, self._prefix + uuid)
+
+        # extract information
+        for rdr, extrs in rdr2ext.items():
+            try:
+                # get content
+                content = rdr(path) if rdr is not None else None
+
+                # apply extractors on this content
+                for ext in extrs:
+                    try:
+                        # get predicate/value tuples
+                        for node, pred, value in ext.extract(subject, content, predicates):
+                            yield node, pred, value
+
+                    except base.errors.ExtractorError as err:
+                        # critical extractor failure.
+                        logger.error('%s failed to extract triples from content: %s', ext, err)
+
+            except base.errors.ReaderError as err:
+                # failed to read any content. skip.
+                logger.error('%s failed to read content: %s', rdr, err)
+
+
+## EOF ##
diff --git a/bsie/utils/bsfs.py b/bsie/utils/bsfs.py
index 01ec5d1..a4b7626 100644
--- a/bsie/utils/bsfs.py
+++ b/bsie/utils/bsfs.py
@@ -10,7 +10,7 @@ import typing
 # bsfs imports
 from bsfs import schema
 from bsfs.namespace import Namespace
-from bsfs.utils import URI, typename
+from bsfs.utils import URI, typename, uuid
 
 # exports
 __all__: typing.Sequence[str] = (
@@ -18,6 +18,7 @@ __all__: typing.Sequence[str] = (
     'URI',
     'schema',
     'typename',
+    'uuid',
     )
 
 ## EOF ##