om.datasets ¶

omegaml.datasets = OmegaStore(bucket=omegaml, prefix=data/)¶

the omegaml.store.base.OmegaStore store for datasets

Methods:

omegaml.store.base.OmegaStore.list()

omegaml.store.base.OmegaStore.get()

omegaml.store.base.OmegaStore.getl()

omegaml.store.base.OmegaStore.put()

omegaml.store.base.OmegaStore.drop()

omegaml.store.base.OmegaStore.exists()

omegaml.store.base.OmegaStore.metadata()

omegaml.store.base.OmegaStore.help()

Mixins:

omegaml.mixins.store.ProjectedMixin

omegaml.mixins.store.LazyGetMixin

omegaml.mixins.store.virtualobj.VirtualObjectMixin

omegaml.mixins.store.promotion.PromotionMixin

omegaml.mixins.mdf.iotools.IOToolsStoreMixin

omegaml.mixins.store.extdmeta.ExtendedMetadataMixin

Backends:

omegaml.backends.npndarray.NumpyNDArrayBackend,

omegaml.backends.virtualobj.VirtualObjectBackend,

omegaml.backends.rawdict.PandasRawDictBackend,

omegaml.backends.rawfiles.PythonRawFileBackend,

Backends ¶

class omegaml.backends.sqlalchemy.SQLAlchemyBackend(model_store=None, data_store=None, tracking=None, **kwargs)¶

sqlalchemy plugin for omegaml

Usage:

Define your sqlalchemy connection:

sqlalchemy_constr = f'sqlalchemy://{user}:{password}@{account}/'

Store the connection in any of three ways:

# -- just the connection
om.datasets.put(sqlalchemy_constr, 'mysqlalchemy')
om.datasets.get('mysqlalchemy', raw=True)
=> the sql connection object

# -- store connection with a predefined sql
om.datasets.put(sqlalchemy_constr, 'mysqlalchemy',
                sql='select ....')
om.datasets.get('mysqlalchemy')
=> will return a pandas dataframe using the specified sql to run.
   specify chunksize= to return an interable of dataframes

# -- predefined sqls can contain variables to be resolved at access time
#    if you miss to specify required variables in sqlvars, a KeyError is raised
om.datasets.put(sqlaclhemy_constr, 'myview',
                sql='select ... from T1 where col="{var}"')
om.datasets.get('mysqlalchemy', sqlvars=dict(var="value"))

# -- Variables are replaced by binding parameters, which is safe for
#    untrusted inputs. To replace variables as strings, use double
#    `{{variable}}` notation. A warning will be issued because this
#    is considered an unsafe practice for untrusted input (REST API).
#    It is in your responsibility to sanitize the value of the `cols` variable.
om.datasets.put(sqlaclhemy_constr, 'myview',
                sql='select {{cols}} from T1 where col="{var}"')
om.datasets.get('mysqlalchemy', sqlvars=dict(cols='foo, bar',
                                             var="value"))

Query data from a connection and store into an omega-ml dataset:

# -- copy the result of the sqlalchemy query to omegaml
om.datasets.put(sqlalchemy_constr, 'mysqlalchemy',
                sql='select ...', copy=True)
om.datasets.get('mysqlalchemy')
=> will return a pandas dataframe (without executing any additional queries)
=> can also use with om.datasets.getl('mysqlalchemy') to return a MDataFrame

Controlling the table used in the connection:

# -- the default table is {bucket}_{name}, override using table='myname'
om.datasets.put(sqlalchemy_constr, 'mysqlalchemy',
                table='mytable',
                sql='select ...',
                copy=True)
om.datasets.get('mysqlalchemy') # read from {bucket}_myname

# -- to use a specific table, without bucket information use table=':myname'
om.datasets.put(sqlalchemy_constr, 'mysqlalchemy',
                table=':mytable',
                sql='select ...',
                copy=True)
om.datasets.get('mysqlalchemy') # read from myname

Inserting data via a previously stored connection:

# -- store data back through the connection
om.datasets.put(sqlalchemy_constr, 'mysqlalchemy')
om.datasets.put(df, 'mysqlalchemy',
                table='SOMETABLE')

Using variables in connection strings:

Connection strings may contain variables, e.g. userid and password. By default variables are resolved from the os environment. Can also specify using any dict.:

# -- use connection string with variables
sqlalchemy_constr = 'sqlite:///{dbname}.db'
om.datasets.put(sqlalchemy_constr, 'userdb')
om.datasets.get('userdb', secrets=dict(dbname='chuckdb'))

# -- alternatively, create a vault dataset:
secrets = dict(userid='chuck', dbname='chuckdb')
om.datasets.put(secrets, '.omega/vault')
om.datasets.get('userdb')

the '.omega/vault' dataset will be queried using the current userid as
the secret name, and the dbname retrieved from the document. This is
experimental and the vault is not encrypted.

Advanced:

om.datasets.put() supports the following additional keyword arguments

chunksize=int - specify the number of rows to read from sqlalchemy in one chunk. defaults to 10000

parse_dates=['col', ...] - list of column names to parse for date, time or datetime. see pd.read_sql for details

transform=callable - a callable, is passed the DataFrame of each chunk before it is inserted into the database. use to provide custom transformations. only works on copy=True

any other kwargs supported by pandas.read_sql

KIND = 'sqlalchemy.conx'¶

get(name, sql=None, chunksize=None, raw=False, sqlvars=None, secrets=None, index=True, keep=None, lazy=False, table=None, trusted=False, *args, **kwargs)¶

retrieve a stored connection or query data from connection

Parameters:

name (str) – the name of the connection
secrets (dict) – dict to resolve variables in the connection string
keep (bool) – if True connection is kept open, defaults to True (change default by setting om.defaults.SQLALCHEMY_ALWAYS_CACHE = False)
table (str) – the name of the table, will be prefixed with the store’s bucket name unless the table is specified as ‘:name’
trusted (bool|str) – if passed must be the value for store.sign(sqlvars or kwargs), otherwise a warning is issued for any remaining variables in the sql statement

Returns:

connection

To query data and return a DataFrame, specify sql='select ...':

Parameters:

sql (str) – the sql query, defaults to the query specific on .put()
chunksize (int) – the number of records for each chunk, if specified returns an iterator
sqlvars (dict) – optional, if specified will be used to format sql

Returns:

pd.DataFrame

To get the connection for a data query, instead of a DataFrame:

Parameters:

raw (bool) – if True, returns the raw sql alchemy connection
keep (bool) – option, if True keeps the connection open. Lazy=True implies keep=True. This is potentially unsafe in a multi-user environment where connection strings contain user-specific secrets. To always keep connections open, set om.datasets.defaults.SQLALCHEMY_ALWAYS_CACHE=True

Returns:

connection

To get a cursor for a data query, instead of a DataFrame. Note this implies keep=True.

Parameters:

lazy (bool) – if True, returns a cursor instead of a DataFrame
sql (str) – the sql query, defaults to the query specific on .put()

Returns:

cursor

put(obj, name, sql=None, copy=False, append=True, chunksize=None, transform=None, table=None, attributes=None, insert=False, secrets=None, *args, **kwargs)¶

store sqlalchemy connection or insert data into an existing connection

Parameters:

obj (str|pd.DataFrame) – the sqlalchemy connection string or a dataframe object
name (str) – the name of the object
table (str) – optional, if specified is stored along connection
sql (str) – optional, if specified is stored along connection
copy (bool) – optional, if True the connection is queried using sql and the resulting data is stored instead, see below
attributes (dict) – optional, set or update metadata.attributes

Returns:

metadata of the stored connection

Instead of inserting the connection specify copy=True to query data and store it as a DataFrame dataset given by name:

Parameters:

sql (str) – sql to query
append (bool) – if True the data is appended if exists already
chunksize (int) – number of records to query in each chunk
transform (callable) – passed as DataFrame.to_sql(method=)

Returns:

metadata of the inserted dataframe

To insert data via a previously stored connection, specify insert=True:

Parameters:

insert (bool) – specify True to insert via the connection
table (str) – the table name to use for inserting data
append (bool) – if False will replace any existing table, defaults to True
index (bool) – if False will not attempt to create an index in target, defaults to False
chunksize (int) – number of records to insert in each chunk

Returns:

metadata of the connection

classmethod supports(obj, name, insert=False, data_store=None, model_store=None, *args, **kwargs)¶: test if this backend supports this obj

class omegaml.backends.npndarray.NumpyNDArrayBackend(model_store=None, data_store=None, tracking=None, **kwargs)¶

Store numpy NDArray of any shape or size

The NDArray is serialized to a byte string and stored as a BLOB. Thus it can have arbitrary size and dimensions, ideal for image data and Tensors.

KIND = 'ndarray.bin'¶

get(name, version=-1, force_python=False, lazy=False, **kwargs)¶

get an obj

Parameters:: name – the name of the object (str)
Returns:: the object as it was originally stored

put(obj, name, attributes=None, allow_pickle=False, **kwargs)¶

put an obj

Parameters:

obj – the object to store (object)
name – the name of the object (str)
attributes – the attributes dict (dict, optional)
kwargs – other kwargs to be passed to the Metadata object

Returns:

the Metadata object

classmethod supports(obj, name, as_pydata=False, **kwargs)¶: test if this backend supports this obj

class omegaml.backends.virtualobj.VirtualObjectBackend(model_store=None, data_store=None, tracking=None, **kwargs)¶

Support arbitrary functions as object handlers

Virtual object functions can be any callable that provides a __omega_virtual__ attribute. The callable must support the following signature:

@virtualobj
def virtualobjfn(data=None, method='get|put|drop',
                 meta=None, store=None, **kwargs):
    ...
    return data

Note that there is a distinction between storing the function as a virtual object, and passing data in or getting data out of the store. It is the responsibility of the function to implement the appropriate code for get, put, drop, as well as to keep track of the data it actually stores.

As a convenience virtual object handlers can be implemented as a subclass of VirtualObjectHandler

Usage:

1) as a virtual data handler

    # create the 'foo' virtual object
    om.datasets.put(virtualobjfn, 'foo')

    # get data from the virtualobj
    om.datasets.get('foo')
    => will call virtualobjfn(method='get')

    # put data into the virtualobj
    om.datasets.put(data, 'foo')
    => will call virtualobjfn(data=data, method='put')

    # drop the virtualfn
    om.datasets.drop('name')
    => will call virtualobjfn(method='drop') and then
       drop the virtual object completely from the storage

2) as a virtual model

    # create the mymodel model as a virtualobj
    om.models.put(virtualobjfn, 'mymodel')

    # run the model's predict() function
    om.runtime.model('mymodel').predict(X)
    => will call virtualobjfn(method='predict')

3) as a virtual script

    # create the myscript script as a virtualobj
    om.models.put(virtualobjfn, 'myscript')

    # run the script
    om.runtime.script('myscript').run()
    => will call virtualobjfn(method='run')

Warning

Virtual objects are executed in the address space of the client or runtime context. Make sure that the source of the code is trustworthy. Note that this is different from Backends and Mixins as these are pro-actively enabled by the administrator of the client or runtime context, respectively - virtual objects can be injected by anyone who are authorized to write data.

KIND = 'virtualobj.dill'¶

get(name, version=-1, force_python=False, lazy=False, **kwargs)¶

get an obj

Parameters:: name – the name of the object (str)
Returns:: the object as it was originally stored

put(obj, name, attributes=None, dill_kwargs=None, as_source=False, **kwargs)¶

put an obj

Parameters:

obj – the object to store (object)
name – the name of the object (str)
attributes – the attributes dict (dict, optional)
kwargs – other kwargs to be passed to the Metadata object

Returns:

the Metadata object

reduce(modelname, results, rName=None, **kwargs)¶

reduce a list of results to a single result

Use this as the last step in a task canvas

Parameters:

modelname (str) – the name of the virtualobj
results (list) – the list of results forwarded by task canvas
rName (result) – the name of the result object
**kwargs

Returns:

result of the virtualobj handler

See Also: om.runtime.mapreduce

classmethod supports(obj, name, **kwargs)¶: test if this backend supports this obj

class omegaml.backends.rawdict.PandasRawDictBackend(model_store=None, data_store=None, tracking=None, **kwargs)¶

OmegaStore backend to support arbitrary collections

Usage:

# store any collection as part of metadata
coll = db['some_collection']
om.datasets.put(coll, 'foo')
=> Metadata(name='foo', collection='some_collection', ...)
# parse the collection using pandas.io.json_normalize
df = om.datasets.get('foo')
# use an alternate parser that accepts dict|list(dict)
df = om.datasets.get('foo', parser=some_fn)
# get a MDataFrame
om.datasets.getl('foo')
# preserve all document keys, including _id
om.datasets.getl('foo', raw=True)

KIND = 'pandas.rawdict'¶

get(name, version=-1, lazy=False, raw=False, parser=None, filter=None, **kwargs)¶

get an obj

Parameters:: name – the name of the object (str)
Returns:: the object as it was originally stored

put(obj, name, attributes=None, as_raw=None, **kwargs)¶

put an obj

Parameters:

obj – the object to store (object)
name – the name of the object (str)
attributes – the attributes dict (dict, optional)
kwargs – other kwargs to be passed to the Metadata object

Returns:

the Metadata object

classmethod supports(obj, name, as_raw=None, data_store=None, **kwargs)¶: test if this backend supports this obj

class omegaml.backends.rawfiles.PythonRawFileBackend(model_store=None, data_store=None, tracking=None, **kwargs)¶

OmegaStore backend to support arbitrary files

KIND = 'python.file'¶

get(name, local=None, mode='wb', open_kwargs=None, **kwargs)¶

get a stored file as a file-like object with binary contents or a local file

Parameters:

name (str) – the name of the file
local (str) – if set the local path will be created and the file stored there. If local does not have an extension it is assumed to be a directory name, in which case the file is stored as the same name.
mode (str) – the mode to use on .open() for the local file
open_kwargs (dict) – the kwargs to use .open() for the local file
**kwargs – any kwargs passed to datasets.metadata()

Returns:

the file-like output handler (local is None) the path to the local file (local is given)

Mixins ¶

class omegaml.mixins.store.ProjectedMixin¶

A OmegaStore mixin to process column specifications in dataset name

get(name, *args, **kwargs)¶

Return a projected dataset given a name of form name[colspec]

colspec can be any of

a comma separated list of columns, e.g. foo[a,b]
an open-ended slice, e.g. foo[a:] => all columns following a, inclusive
an closed slice, e.g. foo[a:b] => all columns between a,b, inclusive
a close-ended slice, e.g. foo[:b] => all columns up to b, inclusive
an empty slice, e.g. foo[:] => all columns
a list of columns to exclude, e.g. foo[^b] => all columns except b

Parameters:: name – (str) the name of the dataset, optionally including a column specification
Returns:: the dataset with projected columns

class omegaml.mixins.store.LazyGetMixin¶

OmegaStore mixin to support chunked lazy get via name

Usage:

equivalent of om.datasets.get('foo', lazy=True).iterchunks():
mdf = om.datasets.get('foo#')
mdf = om.datasets.get('foo#iterchunks')
equivalent of om.datasets.get('foo', lazy=True).iterchunks(chunksize=10):
mdf = om.datasets.get('foo#iterchunks:chunksize=10')
equivalent of om.datasets.get('foo', lazy=True).iloc[0:10]:
mdf = om.datasets.get('foo#rows:start=1,end=10')

class omegaml.mixins.store.virtualobj.VirtualObjectMixin¶

process virtual objects

This checks if an object is a VirtualObject and if so retrieves the handler and processes it.

class omegaml.mixins.store.promotion.PromotionMixin¶

Promote objects from one bucket to another

Promotion Methods:

getput - performs target.put(source.get()) and copies metadata
attributes by merging target.metadata’s .attributes and .kind_meta
metadata - creates a new metadata data entry in target, copying
metadata attributes and kind_meta. Does not get/put the object itself (i.e. no associated data is promoted).
data - like getput, but does not merge metadata
export - performs .to_archive() and .from_archive(), effectively
copying metadata, the associated gridfile (if available) and collection data (if available). This is equivalent to om runtime export.

The default promotion method is getput(), or the object’s backend.PROMOTE method, if specified.

Some object backends provide a default promotion other than getput:

sqlalchemy.conx - uses the metadata promotion, effectively copying only
metadata. Use promote(…, method=’metadata,data’) to also promote data
virtualobj.dill - uses the export promotion, effectively copying all
metadata and the associated @virtualobj function. If the source object is a versioned model, this copies the current version and metadata. To copy a specific version, use promote(‘model@version’). To create a new version in the target bucket use promote(…, method=’getput’).

promote(name, other, asname=None, drop=None, method='default', get=None, put=None, **kwargs)¶

Promote object to another store.

This effectively copies the object. If the objects exists in the target it will be replaced.

Parameters:

name – The name of the object
other – the OmegaStore instance to promote to
asname – the name to use in other, defaults to .metadata(name).name
drop (bool) – if True calls other.drop(force=True) before promoting, defaults to False
method (str|list) – specify the method or multiple methods in sequence, available methods are ‘default’, ‘getput’, ‘metadata’, ‘data’. For ‘default’, the object backend’s .PROMOTE property is used, defaulting to ‘getput’
get (dict) – optional, specifies the store.get(**kwargs)
put (dict) – optional, specifies the other.put(**kwargs)
kwargs – additional kwargs are passed to the initial other.put(), for metadata promotion

Returns:

The Metadata of the new object

class omegaml.mixins.mdf.iotools.IOToolsStoreMixin¶

read_csv(csvfn, name, chunksize=10000, append=False, apply=None, mode='r', open_kwargs=None, **kwargs)¶

read large files from s3, hdfs, http/s, sftp, scp, ssh, write to om.datasets

Usage:

To insert a local csv file into a dataset:
om.datasets.read_csv('/path/filename.csv', 'dataset-name')
To insert a file stored in any of the supported locations specify its fully qualified location and filename. The specific format must be specified according to the smart_open library:
om.datasets.read_csv('https://...', 'dataset-name')
om.datasets.read_csv('s3://...', 'dataset-name')
om.datasets.read_csv('hdfs://...', 'dataset-name')
om.datasets.read_csv('sftp://...', 'dataset-name')
om.datasets.read_csv('scp://...', 'dataset-name')
om.datasets.read_csv('ssh://...', 'dataset-name')
Optionally define a function to receives each chunk as a dataframe and apply further processing (e.g. transformations, filtering):
def process(df):
    # apply any processing to df
    return df

om.datasets.read_csv(...., apply=process)

Parameters:

csvfn (str) – the fully qualified path and name of the csv file, according to the smart_open library
chunksize (int) – the size of each chunk processed before writing to the dataset
append (bool) – if True, appends to the dataset. defaults to False
apply (callable) – if specified, each chunk is forwarded as a DataFrame and the returned result is inserted to the dataset. Use this for transformations or filtering
mode (str) – file open mode, defaults to r
open_kwargs (dict) – additional kwargs to smart_open
**kwargs – additional kwargs are passed to pandas.read_csv

Returns:

MDataFrame

om.datasets¶

Backends¶

Mixins¶

om.datasets ¶

Backends ¶

Mixins ¶