This package provides I/O functionality for skbio.
Reading and writing files (I/O) can be a complicated task:
To address these issues (and others), scikit-bio provides a simple, powerful interface for dealing with I/O. We accomplish this by using a single I/O registry. Below is a description of how to use the registry and how to extend it.
There are two ways to read files. The first way is to use the procedural interface:
my_obj = skbio.io.read(<filehandle or filepath>, format='<format here>', into=<class to construct>)
The second is to use the object-oriented (OO) interface which is automatically constructed from the procedural interface:
my_obj = <class to construct>.read(<filehandle or filepath>, format='<format here>')
For example, to read a newick file using both interfaces you would type:
>>> from skbio import read >>> from skbio import TreeNode >>> from io import StringIO >>> open_filehandle = StringIO(u'(a, b);') >>> tree = read(open_filehandle, format='newick', into=TreeNode) >>> tree <TreeNode, name: unnamed, internal node count: 0, tips count: 2>
For the OO interface:
>>> open_filehandle = StringIO(u'(a, b);') >>> tree = TreeNode.read(open_filehandle, format='newick') >>> tree <TreeNode, name: unnamed, internal node count: 0, tips count: 2>
In the case of skbio.io.read if into is not provided, then a generator will be returned. What the generator yields will depend on what format is being read.
When into is provided, format may be omitted and the registry will use its knowledge of the available formats for the requested class to infer the correct format. This format inference is also available in the OO interface, meaning that format may be omitted there as well.
As an example:
>>> open_filehandle = StringIO(u'(a, b);') >>> tree = TreeNode.read(open_filehandle) >>> tree <TreeNode, name: unnamed, internal node count: 0, tips count: 2>
We call format inference sniffing, much like the csv module of Python’s standard library. The goal of a sniffer is twofold: to identify if a file is a specific format, and if it is, to provide **kwargs which can be used to better parse the file.
There is a built-in sniffer which results in a useful error message if an empty file is provided as input and the format was omitted.
Just as when reading files, there are two ways to write files.
skbio.io.write(my_obj, format='<format here>', into=<filehandle or filepath>)
my_obj.write(<filehandle or filepath>, format='<format here>')
In the procedural interface, format is required. Without it, scikit-bio does not know how you want to serialize an object. OO interfaces define a default format, so it may not be necessary to include it.
For details on what objects are supported by each format, see the associated documentation.
|clustal||Clustal format (skbio.io.clustal)|
|fasta||FASTA/QUAL format (skbio.io.fasta)|
|fastq||FASTQ format (skbio.io.fastq)|
|lsmat||Labeled square matrix format (skbio.io.lsmat)|
|newick||Newick format (skbio.io.newick)|
|ordination||Ordination results format (skbio.io.ordination)|
|phylip||PHYLIP multiple sequence alignment format (skbio.io.phylip)|
|qseq||QSeq format (skbio.io.qseq)|
Formats are considered to be names which represent a way of encoding a file.
|write(obj, format, into[, mode])||Write a supported skbio file format from an instance or a generator.|
|read(fp[, format, into, verify, mode])||Read a supported skbio file format into an instance or a generator.|
|sniff(fp[, cls, mode])||Attempt to guess the format of a file and return format str and kwargs.|
|RecordError||Raised when a record is bad.|
|FieldError||Raised when a field within a record is bad.|
|UnrecognizedFormatError||Raised when a file’s format is unknown, ambiguous, or unidentifiable.|
|FileFormatError||Raised when a file cannot be parsed.|
|ClustalFormatError||Raised when a clustal formatted file cannot be parsed.|
|FASTAFormatError||Raised when a fasta formatted file cannot be parsed.|
|FASTQFormatError||Raised when a fastq formatted file cannot be parsed.|
|LSMatFormatError||Raised when a lsmat formatted file cannot be parsed.|
|NewickFormatError||Raised when a newick formatted file cannot be parsed.|
|OrdinationFormatError||Raised when an ordination formatted file cannot be parsed.|
|PhylipFormatError||Raised when a phylip formatted file cannot be parsed.|
|QSeqFormatError||Raised when a qseq formatted file cannot be parsed.|
To extend I/O in skbio, developers should create a submodule in skbio/io/ named after the file format it implements.
For example, if you were to create readers and writers for a fasta file, you would create a submodule skbio/io/fasta.py. In this submodule you would use the following decorators: register_writer, register_reader, and register_sniffer. These associate your functionality to a format string and potentially an skbio class. Please see the relevant documenation for more information about these functions and the specifications for readers, writers, and sniffers.
Once you are satisfied with the functionality, you will need to ensure that skbio/io/__init__.py contains an import of your new submodule so the decorators are executed on importing the user functions above. Use the function import_module('skbio.io.my_new_format').
The following keyword args may not be used when defining new readers or writers as they already have special meaning to the registry system:
If a keyword argument is a file, such as in the case of fasta with qual, then you can set the default to a specific marker, or sentinel, to indicate to the registry that the kwarg should have special handling. For example:
from skbio.io import FileSentinel @register_reader(fasta, object) def fasta_to_object(fh, qual=FileSentinel): ...
After the registry reads your function, it will replace FileSentinel with None allowing you to perform normal checks for kwargs (e.g. if my_kwarg is not None:). If a user provides input for the kwarg, the registry will convert it to an open filehandle.
Keyword arguments are not permitted in sniffers. Sniffers may not raise exceptions; if an exception is thrown by a sniffer, the user will be asked to report it on our issue tracker.
When raising errors in readers and writers, the error should be a subclass of FileFormatError specific to your new format.
Because scikit-bio handles all of the I/O boilerplate, you only need to test the actual business logic of your readers, writers, and sniffers. The easiest way to accomplish this is to create a list of files and their expected results when deserialized. Then you can iterate through the list ensuring the expected results occur and that the expected results can be reserialized into an equivalent file. This process is called ‘roundtripping’.
It is also important to test some invalid inputs and ensure that the correct error is raised by your readers. Consider using assertRaises as a context manager like so:
with self.assertRaises(SomeFileFormatErrorSubclass) as cm: do_something_wrong() self.assertIn('action verb or subject of an error', str(cm.exception))
A good example to review when preparing to write your first I/O unit tests is the ordination test code (see in skbio/io/tests/test_ordination.py).
|register_writer(format[, cls])||Return a decorator for a writer function.|
|register_reader(format[, cls])||Return a decorator for a reader function.|
|register_sniffer(format)||Return a decorator for a sniffer function.|
|list_write_formats(cls)||Return a list of available write formats for a given cls instance.|
|list_read_formats(cls)||Return a list of available read formats for a given cls type.|
|get_writer(format[, cls])||Return a writer for a format.|
|get_reader(format[, cls])||Return a reader for a format.|
|get_sniffer(format)||Return a sniffer for a format.|