Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix the nx_char type for numpy to and . #554

Open
wants to merge 7 commits into
base: master
Choose a base branch
from
Open
Show file tree
Hide file tree
Changes from 4 commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
101 changes: 63 additions & 38 deletions src/pynxtools/dataconverter/helpers.py
Original file line number Diff line number Diff line change
Expand Up @@ -591,71 +591,93 @@ def is_value_valid_element_of_enum(value, elist) -> Tuple[bool, list]:
np.uint16,
np.uint32,
np.uint64,
np.uint,
np.unsignedinteger,
np.signedinteger,
)
np_float = (np.float16, np.float32, np.float64, np.floating)
np_bytes = (np.bytes_, np.byte, np.ubyte)
np_char = (np.str_, np.char.chararray, *np_bytes)
# Not to be confused with `np.byte` and `np.ubyte`, these store
# and integer of `8bit` and `unsigned 8bit` respectively.
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
# Not to be confused with `np.byte` and `np.ubyte`, these store
# and integer of `8bit` and `unsigned 8bit` respectively.
# Not to be confused with `np.byte` and `np.ubyte`, these store
# integers of `8bit` and `unsigned 8bit` respectively.

What exactly is not to be confused? Maybe write "np.xxx is not to be confused..."

np_bytes = (np.bytes_,)
np_char = (np.str_, np.bytes_) # Only numpy Unicode string and Byte string
np_bool = (np.bool_,)
np_complex = (np.complex64, np.complex128, np.cdouble, np.csingle)
NEXUS_TO_PYTHON_DATA_TYPES = {
"ISO8601": (str,),
"NX_BINARY": (
bytes,
bytearray,
np.ndarray,
*np_bytes,
),
"NX_BOOLEAN": (bool, np.ndarray, *np_bool),
"NX_CHAR": (str, np.ndarray, *np_char),
"NX_BOOLEAN": (bool, *np_bool),
"NX_CHAR": (str, *np_char),
"NX_DATE_TIME": (str,),
"NX_FLOAT": (float, np.ndarray, *np_float),
"NX_INT": (int, np.ndarray, *np_int),
"NX_UINT": (np.ndarray, np.unsignedinteger),
"NX_FLOAT": (float, *np_float),
"NX_INT": (int, *np_int),
"NX_UINT": (
np.unsignedinteger,
np.uint,
),
"NX_NUMBER": (
int,
float,
np.ndarray,
*np_int,
*np_float,
dict,
),
"NX_POSINT": (
int,
np.ndarray,
np.signedinteger,
), # > 0 is checked in is_valid_data_field()
"NX_COMPLEX": (complex, np.ndarray, *np_complex),
"NXDL_TYPE_UNAVAILABLE": (str,), # Defaults to a string if a type is not provided.
"NX_COMPLEX": (complex, *np_complex),
"NXDL_TYPE_UNAVAILABLE": (
str,
*np_char,
), # Defaults to a string if a type is not provided.
"NX_CHAR_OR_NUMBER": (
str,
int,
float,
np.ndarray,
*np_char,
*np_int,
*np_float,
dict,
),
}


def check_all_children_for_callable(objects: list, check: Callable, *args) -> bool:
"""Checks whether all objects in list are validated by given callable."""
for obj in objects:
if not check(obj, *args):
return False
def check_all_children_for_callable(
objects: Union[list, np.ndarray],
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We should extend this to all iterable data types (e.g. tuples, ...)

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

All the datastructure are not officially mentioned in NeXus datatype. Only allowd datastructures are array or list which is equivalent to array. But here I added check for Tuple, I will not update the parameter annotation as we are not expecting that.

checker: Optional[Callable] = None,
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I suggest to rename this to callable

accepted_types: Optional[tuple] = None,
) -> bool:
"""Checks whether all objects in list or numpy array are validated
by given callable and types.
"""

return True
if checker is not None:
for obj in objects:
args = (obj, accepted_types) if accepted_types is not None else (obj,)
if not checker(*args):
return False
return True

# default checker
tmp_arr = None
if isinstance(objects, list):
# Handles list and list of list
tmp_arr = np.array(objects)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Will this work of the dtypes within a list are not the same?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think it will work. Because, numpy will check if all the elments have the same size of bit atleast converable to a common datatype, Otherwise it fails. If there is heterogeneous datatype (like string, nemeric) numpy will choose unicode U/ python O or something like that.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ahh, it will not work in specificatino of NX_FLOAT or NX_INT

elif isinstance(objects, np.ndarray):
tmp_arr = objects
if tmp_arr is not None:
return any([np.issubdtype(tmp_arr.dtype, type_) for type_ in accepted_types])
return False


def is_valid_data_type(value, accepted_types):
"""Checks whether the given value or its children are of an accepted type."""
if not isinstance(value, list):

if not isinstance(value, (list, np.ndarray)):
return isinstance(value, accepted_types)

return check_all_children_for_callable(value, isinstance, accepted_types)
return check_all_children_for_callable(objects=value, accepted_types=accepted_types)


def is_positive_int(value):
Expand All @@ -665,7 +687,7 @@ def is_greater_than(num):
return num.flat[0] > 0 if isinstance(num, np.ndarray) else num > 0

if isinstance(value, list):
return check_all_children_for_callable(value, is_greater_than)
return check_all_children_for_callable(objects=value, checker=is_greater_than)

return value.flat[0] > 0 if isinstance(value, np.ndarray) else value > 0

Expand All @@ -685,28 +707,31 @@ def convert_str_to_bool_safe(value):
def is_valid_data_field(value, nxdl_type, path):
# todo: Check this funciton and wtire test for it. It seems the funciton is not
# working as expected.
"""Checks whether a given value is valid according to what is defined in the NXDL.
"""Checks whether a given value is valid according to the type defined in the NXDL.

This function will also try to convert typical types, for example int to float,
and return the successful conversion.
This function also converts bool value comes in str format. In case, it fails to
convert, it raises an Exception.
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
This function also converts bool value comes in str format. In case, it fails to
convert, it raises an Exception.
This function also converts boolean value that are given as strings (i.e., "True" to True).

It doesn't really raise an Exception, but just a ValidationProblem.InvalidDatetime warning. What were you trying to say here?

Copy link
Collaborator Author

@RubelMozumder RubelMozumder Feb 27, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I do not see the InvalidDateTime warning anywhere.


If it fails to convert, it raises an Exception.

Returns two values: first, boolean (True if the the value corresponds to nxdl_type,
False otherwise) and second, result of attempted conversion or the original value
(if conversion is not needed or impossible)
Returns two values:
boolean (True if the the value corresponds to nxdl_type, False otherwise)
converted_value bool value.
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Add typing annotation. The calling function expects a single bool.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The return handling is correct, nevertheless I suggest to add typing everywhere.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Doc is updated. You might be talking about the return is_valid_data_field(mapping[key], node.dtype, key)[0] in `validation.py. It returns a single first value. Extra variable is added.

"""
accepted_types = NEXUS_TO_PYTHON_DATA_TYPES[nxdl_type]
output_value = value

accepted_types = NEXUS_TO_PYTHON_DATA_TYPES[nxdl_type]
# Do not count the dict as it represents a link value
if not isinstance(value, dict) and not is_valid_data_type(value, accepted_types):
try:
if accepted_types[0] is bool and isinstance(value, str):
value = convert_str_to_bool_safe(value)
if value is None:
raise ValueError
output_value = accepted_types[0](value)
except ValueError:
return True, value

collector.collect_and_log(
path, ValidationProblem.InvalidType, accepted_types, nxdl_type
)
return False, value
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This return value False here is interpreted as "undocumented field" in the calling function. So we get an additional wrong warning if the dtype does not match.

Copy link
Collaborator Author

@RubelMozumder RubelMozumder Feb 27, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I see that it is just for internal logic (please see line 750-766 in validation.py). In is_docmented it checks the datatype, node type (is field or group). But the real type of the errors are collect and print according to the validation error type in and from collect function.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't understand what you are trying to say here. My point is that is_documented returns False if the datatype does not match, which I would say is wrong, because it is documented, but just has the wrong data type.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This behavior was btw. introduced in #522, I believe. Maybe @GinzburgLev or @lukaspie can comment?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I wanted to explain the code a bit, to clarify that in that code, no different type validation errors are collected, as you are assuming undocumented field warning will pop up.
By the way, I removed the second value from return statement as it is not being used anywhere.

except (ValueError, TypeError):
collector.collect_and_log(
path, ValidationProblem.InvalidType, accepted_types, nxdl_type
)
Expand All @@ -726,7 +751,7 @@ def is_valid_data_field(value, nxdl_type, path):
collector.collect_and_log(path, ValidationProblem.InvalidDatetime, value)
return False, value

return True, output_value
return True, value


@lru_cache(maxsize=None)
Expand Down
4 changes: 2 additions & 2 deletions src/pynxtools/dataconverter/validation.py
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Adjust all occurences of is_valid_data_field (e.g. line 552)

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@RubelMozumder @sanbrock Returning False in L552 if the data type of a field does not fit effectively makes this field "undocumented". Is this intentional? The produced warnings at least are rather contradictive:

WARNING: The value at /ENTRY[entry]/data/@axes should be one of: (<class 'str'>, <class 'numpy.str_'>, <class 'numpy.bytes_'>), as defined in the NXDL as NX_CHAR.
WARNING: Field /ENTRY[entry]/data/@axes written without documentation.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@sanbrock @sherjeelshabih Also, apparently during this second round of checking for undocumented fields, that are not being found before, no checks for enumerations are being done. I don't really understand the logic of this checking completely, but I suggest a checking for enums should happen here as well.
Why can't we re-use the mechanism in recurse_tree here, i.e. the handling_map and corresponding functions?

Original file line number Diff line number Diff line change
Expand Up @@ -422,7 +422,7 @@ def handle_field(node: NexusNode, keys: Mapping[str, Any], prev_path: str):
continue

# Check general validity
is_valid_data_field(
_, _ = is_valid_data_field(
mapping[f"{prev_path}/{variant}"], node.dtype, f"{prev_path}/{variant}"
)

Expand Down Expand Up @@ -468,7 +468,7 @@ def handle_attribute(node: NexusNode, keys: Mapping[str, Any], prev_path: str):
return

for variant in variants:
is_valid_data_field(
_, _ = is_valid_data_field(
mapping[
f"{prev_path}/{variant if variant.startswith('@') else f'@{variant}'}"
],
Expand Down
Loading
Loading