Definitions

class odps.ODPS(access_id=None, secret_access_key=None, project=None, endpoint=None, **kw)[source]

Main entrance to ODPS.

Convenient operations on ODPS objects are provided. Please refer to ODPS docs to see the details.

Generally, basic operations such as list, get, exist, create, delete are provided for each ODPS object. Take the Table as an example.

To create an ODPS instance, access_id and access_key is required, and should ensure correctness, or SignatureNotMatch error will throw. If tunnel_endpoint is not set, the tunnel API will route service URL automatically.

Parameters:
  • access_id – Aliyun Access ID
  • secret_access_key – Aliyun Access Key
  • project – default project name
  • endpoint – Rest service URL
  • tunnel_endpoint – Tunnel service URL
  • logview_host – Logview host URL
Example:
>>> odps = ODPS('**your access id**', '**your access key**', 'default_project')
>>>
>>> for table in odps.list_tables():
>>>    # handle each table
>>>
>>> table = odps.get_table('dual')
>>>
>>> odps.exist_table('dual') is True
>>>
>>> odps.create_table('test_table', schema)
>>>
>>> odps.delete_table('test_table')
attach_session(session_name, taskname=None, hints=None)[source]

Attach to an existing session.

Parameters:
  • session_name – The session name.
  • taskname – The created sqlrt task name. If not provided, the default value is used. Mostly doesn’t matter, default works.
Returns:

A SessionInstance you may execute select tasks within.

copy_offline_model(name, new_name, project=None, new_project=None, async_=False)[source]

Copy current model into a new location.

Parameters:
  • new_name – name of the new model
  • new_project – new project name. if absent, original project name will be used
  • async – if True, return the copy instance. otherwise return the newly-copied model
create_fs_volume(name, project=None, **kwargs)[source]

Create a new-fashioned file system volume in a project.

Parameters:
  • name (str) – volume name name
  • project (str) – project name, if not provided, will be the default project
Returns:

volume

Return type:

odps.models.FSVolume

See also

odps.models.FSVolume

create_function(name, project=None, **kwargs)[source]

Create a function by given name.

Parameters:
  • name – function name
  • project – project name, if not provided, will be the default project
  • class_type (str) – main class
  • resources (list) – the resources that function needs to use
Returns:

the created function

Return type:

odps.models.Function

Example:
>>> res = odps.get_resource('test_func.py')
>>> func = odps.create_function('test_func', class_type='test_func.Test', resources=[res, ])
create_parted_volume(name, project=None, **kwargs)[source]

Create an old-fashioned partitioned volume in a project.

Parameters:
  • name (str) – volume name name
  • project (str) – project name, if not provided, will be the default project
Returns:

volume

Return type:

odps.models.PartedVolume

See also

odps.models.PartedVolume

create_resource(name, type=None, project=None, **kwargs)[source]

Create a resource by given name and given type.

Currently, the resource type can be file, jar, py, archive, table.

The file, jar, py, archive can be classified into file resource. To init the file resource, you have to provide another parameter which is a file-like object.

For the table resource, the table name, project name, and partition should be provided which the partition is optional.

Parameters:
  • name – resource name
  • type – resource type, now support file, jar, py, archive, table
  • project – project name, if not provided, will be the default project
  • kwargs – optional arguments, I will illustrate this in the example below.
Returns:

resource depends on the type, if file will be odps.models.FileResource and so on

Return type:

odps.models.Resource’s subclasses

Example:
>>> from odps.models.resource import *
>>>
>>> res = odps.create_resource('test_file_resource', 'file', fileobj=open('/to/path/file'))
>>> assert isinstance(res, FileResource)
>>> True
>>>
>>> res = odps.create_resource('test_py_resource.py', 'py', fileobj=StringIO('import this'))
>>> assert isinstance(res, PyResource)
>>> True
>>>
>>> res = odps.create_resource('test_table_resource', 'table', table_name='test_table', partition='pt=test')
>>> assert isinstance(res, TableResource)
>>> True
>>>
create_role(name, project=None)[source]

Create a role in a project

Parameters:
  • name – name of the role to create
  • project – project name, if not provided, will be the default project
Returns:

role object created

create_session(session_worker_count, session_worker_memory, session_name=None, worker_spare_span=None, taskname=None, hints=None)[source]

Create session.

Parameters:
  • session_worker_count – How much workers assigned to the session.
  • session_worker_memory – How much memory each worker consumes.
  • session_name – The session name. Not specifying to use its ID as name.
  • worker_spare_span – format “00-24”, allocated workers will be reduced during this time. Not specifying to disable this.
  • taskname – The created sqlrt task name. If not provided, the default value is used. Mostly doesn’t matter, default works.
  • hints – Extra hints provided to the session. Parameters of this method will override certain hints.
Returns:

A SessionInstance you may execute select tasks within.

create_table(name, schema, project=None, comment=None, if_not_exists=False, lifecycle=None, shard_num=None, hub_lifecycle=None, async_=False, **kw)[source]

Create an table by given schema and other optional parameters.

Parameters:
  • name – table name
  • schema – table schema. Can be an instance of odps.models.Schema or a string like ‘col1 string, col2 bigint’
  • project – project name, if not provided, will be the default project
  • comment – table comment
  • if_not_exists (bool) – will not create if this table already exists, default False
  • lifecycle (int) – table’s lifecycle. If absent, options.lifecycle will be used.
  • shard_num (int) – table’s shard num
  • hub_lifecycle (int) – hub lifecycle
  • async (bool) – if True, will run asynchronously
Returns:

the created Table if not async else odps instance

Return type:

odps.models.Table or odps.models.Instance

create_user(name, project=None)[source]

Add a user into the project

Parameters:
  • name – user name
  • project – project name, if not provided, will be the default project
Returns:

user created

create_volume_directory(volume, path=None, project=None)[source]

Create a directory under a file system volume.

Parameters:
  • volume (str) – name of the volume.
  • path (str) – path of the directory to be created.
  • project (str) – project name, if not provided, will be the default project.
Returns:

directory object.

default_session()[source]

Attach to the default session of your project.

Returns:A SessionInstance you may execute select tasks within.
delete_function(name, project=None)[source]

Delete a function by given name.

Parameters:
  • name – function name
  • project – project name, if not provided, will be the default project
Returns:

None

delete_offline_model(name, project=None, if_exists=False)[source]

Delete the offline model by given name.

Parameters:
  • name – offline model’s name
  • if_exists – will not raise errors when the offline model does not exist, default False
  • project – project name, if not provided, will be the default project
Returns:

None

delete_resource(name, project=None)[source]

Delete resource by given name.

Parameters:
  • name – resource name
  • project – project name, if not provided, will be the default project
Returns:

None

delete_role(name, project=None)[source]

Delete a role in a project

Parameters:
  • name – name of the role to delete
  • project – project name, if not provided, will be the default project
delete_table(name, project=None, if_exists=False, async_=False, **kw)[source]

Delete the table with given name

Parameters:
  • name – table name
  • project – project name, if not provided, will be the default project
  • if_exists – will not raise errors when the table does not exist, default False
  • async (bool) – if True, will run asynchronously
Returns:

None if not async else odps instance

delete_user(name, project=None)[source]

Delete a user from the project

Parameters:
  • name – user name
  • project – project name, if not provided, will be the default project
delete_volume(name, project=None)[source]

Delete volume by given name.

Parameters:
  • name – volume name
  • project – project name, if not provided, will be the default project
Returns:

None

delete_volume_file(volume, path=None, recursive=False, project=None)[source]

Delete a file / directory object under a file system volume.

Parameters:
  • volume (str) – name of the volume.
  • path (str) – path of the directory to be created.
  • recursive (bool) – if True, recursively delete files
  • project (str) – project name, if not provided, will be the default project.
Returns:

directory object.

delete_volume_partition(volume, partition=None, project=None)[source]

Delete partition in a volume by given name

Parameters:
  • volume (str) – volume name
  • partition (str) – partition name
  • project (str) – project name, if not provided, will be the default project
delete_xflow(name, project=None)[source]

Delete xflow by given name.

Parameters:
  • name – xflow name
  • project – project name, if not provided, will be the default project
Returns:

None

execute_archive_table(table, partition=None, project=None, hints=None, priority=None)[source]

Execute a task to archive tables and wait for termination.

Parameters:
  • table – name of the table to archive
  • partition – partition to archive
  • project – project name, if not provided, will be the default project
  • hints – settings for table archive task.
  • priority – instance priority, 9 as default
Returns:

instance

Return type:

odps.models.Instance

execute_merge_files(table, partition=None, project=None, hints=None, priority=None)[source]

Execute a task to merge multiple files in tables and wait for termination.

Parameters:
  • table – name of the table to optimize
  • partition – partition to optimize
  • project – project name, if not provided, will be the default project
  • hints – settings for merge task.
  • priority – instance priority, 9 as default
Returns:

instance

Return type:

odps.models.Instance

execute_sql(sql, project=None, priority=None, running_cluster=None, hints=None, **kwargs)[source]

Run a given SQL statement and block until the SQL executed successfully.

Parameters:
  • sql (str) – SQL statement
  • project – project name, if not provided, will be the default project
  • priority (int) – instance priority, 9 as default
  • running_cluster – cluster to run this instance
  • hints (dict) – settings for SQL, e.g. odps.mapred.map.split.size
Returns:

instance

Return type:

odps.models.Instance

Example:
>>> instance = odps.execute_sql('select * from dual')
>>> with instance.open_reader() as reader:
>>>     for record in reader:  # iterate to handle result with schema
>>>         # handle each record
>>>
>>> instance = odps.execute_sql('desc dual')
>>> with instance.open_reader() as reader:
>>>     print(reader.raw)  # without schema, just get the raw result
execute_sql_cost(sql, project=None, hints=None, **kwargs)[source]
Parameters:
  • sql (str) – SQL statement
  • project – project name, if not provided, will be the default project
  • hints (dict) – settings for SQL, e.g. odps.mapred.map.split.size
Returns:

cost info in dict format

Return type:

cost: dict

Example:
>>> sql_cost = odps.execute_sql_cost('select * from dual')
>>> sql_cost.udf_num
0
>>> sql_cost.complexity
1.0
>>> sql_cost.input_size
100
execute_xflow(xflow_name, xflow_project=None, parameters=None, project=None, hints=None, priority=None)[source]

Run xflow by given name, xflow project, paremeters, block until xflow executed successfully.

Parameters:
  • xflow_name (str) – XFlow name
  • xflow_project (str) – the project XFlow deploys
  • parameters (dict) – parameters
  • project – project name, if not provided, will be the default project
  • hints (dict) – execution hints
  • priority (int) – instance priority, 9 as default
Returns:

instance

Return type:

odps.models.Instance

exist_function(name, project=None)[source]

If the function with given name exists or not.

Parameters:
  • name – function name
  • project – project name, if not provided, will be the default project
Returns:

True if the function exists or False

Return type:

bool

exist_instance(id_, project=None)[source]

If the instance with given id exists or not.

Parameters:
  • id – instance id
  • project – project name, if not provided, will be the default project
Returns:

True if exists or False

Return type:

bool

exist_offline_model(name, project=None)[source]

If the offline model with given name exists or not.

Parameters:
  • name – offline model’s name
  • project – project name, if not provided, will be the default project
Returns:

True if offline model exists else False

Return type:

bool

exist_project(name)[source]

If project name which provided exists or not.

Parameters:name – project name, if not provided, will be the default project
Returns:True if exists or False
Return type:bool
exist_resource(name, project=None)[source]

If the resource with given name exists or not.

Parameters:
  • name – resource name
  • project – project name, if not provided, will be the default project
Returns:

True if exists or False

Return type:

bool

exist_role(name, project=None)[source]

Check if a role exists in a project

Parameters:
  • name – name of the role
  • project – project name, if not provided, will be the default project
exist_table(name, project=None)[source]

If the table with given name exists or not.

Parameters:
  • name – table name
  • project – project name, if not provided, will be the default project
Returns:

True if table exists or False

Return type:

bool

exist_user(name, project=None)[source]

Check if a user exists in the project

Parameters:
  • name – user name
  • project – project name, if not provided, will be the default project
exist_volume(name, project=None)[source]

If the volume with given name exists or not.

Parameters:
  • name (str) – volume name
  • project (str) – project name, if not provided, will be the default project
Returns:

True if exists or False

Return type:

bool

exist_volume_partition(volume, partition=None, project=None)[source]

If the volume with given name exists in a partition or not.

Parameters:
  • volume (str) – volume name
  • partition (str) – partition name
  • project (str) – project name, if not provided, will be the default project
exist_xflow(name, project=None)[source]

If the xflow with given name exists or not.

Parameters:
  • name – xflow name
  • project – project name, if not provided, will be the default project
Returns:

True if exists or False

Return type:

bool

get_function(name, project=None)[source]

Get the function by given name

Parameters:
  • name – function name
  • project – project name, if not provided, will be the default project
Returns:

the right function

Raise:

odps.errors.NoSuchObject if not exists

get_instance(id_, project=None)[source]

Get instance by given instance id.

Parameters:
  • id – instance id
  • project – project name, if not provided, will be the default project
Returns:

the right instance

Return type:

odps.models.Instance

Raise:

odps.errors.NoSuchObject if not exists

get_logview_address(instance_id, hours=None, project=None)[source]

Get logview address by given instance id and hours.

Parameters:
  • instance_id – instance id
  • hours
  • project – project name, if not provided, will be the default project
Returns:

logview address

Return type:

str

get_logview_host()[source]

Get logview host address. :return: logview host address

get_offline_model(name, project=None)[source]

Get offline model by given name

Parameters:
  • name – offline model name
  • project – project name, if not provided, will be the default project
Returns:

offline model

Return type:

odps.models.ml.OfflineModel

Raise:

odps.errors.NoSuchObject if not exists

get_project(name=None)[source]

Get project by given name.

Parameters:name – project name, if not provided, will be the default project
Returns:the right project
Return type:odps.models.Project
Raise:odps.errors.NoSuchObject if not exists
get_project_policy(project=None)[source]

Get policy of a project

Parameters:project – project name, if not provided, will be the default project
Returns:JSON object
get_resource(name, project=None)[source]

Get a resource by given name

Parameters:
  • name – resource name
  • project – project name, if not provided, will be the default project
Returns:

the right resource

Return type:

odps.models.Resource

Raise:

odps.errors.NoSuchObject if not exists

get_role_policy(name, project=None)[source]

Get policy object of a role

Parameters:
  • name – name of the role
  • project – project name, if not provided, will be the default project
Returns:

JSON object

get_security_option(option_name, project=None)[source]

Get one security option of a project

Parameters:
  • option_name – name of the security option. Please refer to ODPS options for more details.
  • project – project name, if not provided, will be the default project
Returns:

option value

get_security_options(project=None)[source]

Get all security options of a project

Parameters:project – project name, if not provided, will be the default project
Returns:SecurityConfiguration object
get_table(name, project=None)[source]

Get table by given name.

Parameters:
  • name – table name
  • project – project name, if not provided, will be the default project
Returns:

the right table

Return type:

odps.models.Table

Raise:

odps.errors.NoSuchObject if not exists

get_volume(name, project=None)[source]

Get volume by given name.

Parameters:
  • name (str) – volume name
  • project (str) – project name, if not provided, will be the default project
Returns:

volume object. Return type depends on the type of the volume.

Return type:

odps.models.Volume

get_volume_file(volume, path=None, project=None)[source]

Get a file under a partition of a parted volume, or a file / directory object under a file system volume.

Parameters:
  • volume (str) – name of the volume.
  • path (str) – path of the directory to be created.
  • project (str) – project name, if not provided, will be the default project.
Returns:

directory object.

get_volume_partition(volume, partition=None, project=None)[source]

Get partition in a parted volume by given name.

Parameters:
  • volume (str) – volume name
  • partition (str) – partition name
  • project (str) – project name, if not provided, will be the default project
Returns:

partitions

Return type:

odps.models.VolumePartition

get_xflow(name, project=None)[source]

Get xflow by given name

Parameters:
  • name – xflow name
  • project – project name, if not provided, will be the default project
Returns:

xflow

Return type:

odps.models.XFlow

Raise:

odps.errors.NoSuchObject if not exists

See also

odps.models.XFlow

get_xflow_results(instance, project=None)[source]

The result given the results of xflow

Parameters:
  • instance (odps.models.Instance) – instance of xflow
  • project – project name, if not provided, will be the default project
Returns:

xflow result

Return type:

dict

get_xflow_sub_instances(instance, project=None)[source]

The result iterates the sub instance of xflow

Parameters:
  • instance (odps.models.Instance) – instance of xflow
  • project – project name, if not provided, will be the default project
Returns:

sub instances dictionary

iter_xflow_sub_instances(instance, interval=1, project=None)[source]

The result iterates the sub instance of xflow and will wait till instance finish

Parameters:
  • instance (odps.models.Instance) – instance of xflow
  • interval – time interval to check
  • project – project name, if not provided, will be the default project
Returns:

sub instances dictionary

list_functions(project=None)[source]

List all functions of a project.

Parameters:project – project name, if not provided, will be the default project
Returns:functions
Return type:generator
list_instance_queueing_infos(project=None, status=None, only_owner=None, quota_index=None)[source]

List instance queueing information.

Parameters:
  • project – project name, if not provided, will be the default project
  • status – including ‘Running’, ‘Suspended’, ‘Terminated’
  • only_owner (bool) – True will filter the instances created by current user
  • quota_index (str) –
Returns:

instance queueing infos

Return type:

list

list_instances(project=None, start_time=None, end_time=None, status=None, only_owner=None, quota_index=None, **kw)[source]

List instances of a project by given optional conditions including start time, end time, status and if only the owner.

Parameters:
  • project – project name, if not provided, will be the default project
  • start_time (datetime, int or float) – the start time of filtered instances
  • end_time (datetime, int or float) – the end time of filtered instances
  • status – including ‘Running’, ‘Suspended’, ‘Terminated’
  • only_owner (bool) – True will filter the instances created by current user
  • quota_index (str) –
Returns:

instances

Return type:

list

list_offline_models(project=None, prefix=None, owner=None)[source]

List offline models of project by optional filter conditions including prefix and owner.

Parameters:
  • project – project name, if not provided, will be the default project
  • prefix – prefix of offline model’s name
  • owner – Aliyun account
Returns:

offline models

Return type:

list

list_projects(owner=None, user=None, group=None, prefix=None, max_items=None)[source]

List projects.

Parameters:
  • owner – Aliyun account, the owner which listed projects belong to
  • user – name of the user who has access to listed projects
  • group – name of the group listed projects belong to
  • prefix – prefix of names of listed projects
  • max_items – the maximal size of result set
Returns:

projects in this endpoint.

Return type:

generator

list_resources(project=None)[source]

List all resources of a project.

Parameters:project – project name, if not provided, will be the default project
Returns:resources
Return type:generator
list_role_users(name, project=None)[source]

List users who have the specified role.

Parameters:
  • name – name of the role
  • project – project name, if not provided, will be the default project
Returns:

collection of User objects

list_roles(project=None)[source]

List all roles in a project

Parameters:project – project name, if not provided, will be the default project
Returns:collection of role objects
list_tables(project=None, prefix=None, owner=None)[source]

List all tables of a project. If prefix is provided, the listed tables will all start with this prefix. If owner is provided, the listed tables will be belong to such owner.

Parameters:
  • project – project name, if not provided, will be the default project
  • prefix (str) – the listed tables start with this prefix
  • owner (str) – Aliyun account, the owner which listed tables belong to
Returns:

tables in this project, filtered by the optional prefix and owner.

Return type:

generator

list_tables_model(prefix='', project=None)

List all TablesModel in the given project.

Parameters:
  • prefix – model prefix
  • project (str) – project name, if you want to look up in another project
Return type:

list[str]

list_user_roles(name, project=None)[source]

List roles of the specified user

Parameters:
  • name – user name
  • project – project name, if not provided, will be the default project
Returns:

collection of Role object

list_users(project=None)[source]

List users in the project

Parameters:project – project name, if not provided, will be the default project
Returns:collection of User objects
list_volume_files(volume, partition=None, project=None)[source]

List files in a volume. In partitioned volumes, the function returns files under specified partition. In file system volumes, the function returns files under specified path.

Parameters:
  • volume (str) – volume name
  • partition (str) – partition name for partitioned volumes, and path for file system volumes.
  • project (str) – project name, if not provided, will be the default project
Returns:

files

Return type:

list

Example:
>>> # List files under a partition in a partitioned volume. Two calls are equivalent.
>>> odps.list_volume_files('parted_volume', 'partition_name')
>>> odps.list_volume_files('/parted_volume/partition_name')
>>> # List files under a path in a file system volume. Two calls are equivalent.
>>> odps.list_volume_files('fs_volume', 'dir1/dir2')
>>> odps.list_volume_files('/fs_volume/dir1/dir2')
list_volume_partitions(volume, project=None)[source]

List partitions of a volume.

Parameters:
  • volume (str) – volume name
  • project (str) – project name, if not provided, will be the default project
Returns:

partitions

Return type:

list

list_volumes(project=None, owner=None)[source]

List volumes of a project.

Parameters:
  • project – project name, if not provided, will be the default project
  • owner (str) – Aliyun account
Returns:

volumes

Return type:

list

list_xflows(project=None, owner=None)[source]

List xflows of a project which can be filtered by the xflow owner.

Parameters:
  • project – project name, if not provided, will be the default project
  • owner (str) – Aliyun account
Returns:

xflows

Return type:

list

move_volume_file(old_path, new_path, replication=None, project=None)[source]

Move a file / directory object under a file system volume to another location in the same volume.

Parameters:
  • old_path (str) – old path of the volume file.
  • new_path (str) – target path of the moved file.
  • replication (int) – file replication.
  • project (str) – project name, if not provided, will be the default project.
Returns:

directory object.

open_resource(name, project=None, mode='r+', encoding='utf-8')[source]

Open a file resource as file-like object. This is an elegant and pythonic way to handle file resource.

The argument mode stands for the open mode for this file resource. It can be binary mode if the ‘b’ is inside. For instance, ‘rb’ means opening the resource as read binary mode while ‘r+b’ means opening the resource as read+write binary mode. This is most import when the file is actually binary such as tar or jpeg file, so be aware of opening this file as a correct mode.

Basically, the text mode can be ‘r’, ‘w’, ‘a’, ‘r+’, ‘w+’, ‘a+’ just like the builtin python open method.

  • r means read only
  • w means write only, the file will be truncated when opening
  • a means append only
  • r+ means read+write without constraint
  • w+ will truncate first then opening into read+write
  • a+ can read+write, however the written content can only be appended to the end
Parameters:
  • name (odps.models.FileResource or str) – file resource or file resource name
  • project – project name, if not provided, will be the default project
  • mode – the mode of opening file, described as above
  • encoding – utf-8 as default
Returns:

file-like object

Example:
>>> with odps.open_resource('test_resource', 'r') as fp:
>>>     fp.read(1)  # read one unicode character
>>>     fp.write('test')  # wrong, cannot write under read mode
>>>
>>> with odps.open_resource('test_resource', 'wb') as fp:
>>>     fp.readlines() # wrong, cannot read under write mode
>>>     fp.write('hello world') # write bytes
>>>
>>> with odps.open_resource('test_resource') as fp: # default as read-write mode
>>>     fp.seek(5)
>>>     fp.truncate()
>>>     fp.flush()
open_volume_reader(volume, partition=None, file_name=None, project=None, start=None, length=None, **kwargs)[source]

Open a volume file for read. A file-like object will be returned which can be used to read contents from volume files.

Parameters:
  • volume (str) – name of the volume
  • partition (str) – name of the partition
  • file_name (str) – name of the file
  • project (str) – project name, if not provided, will be the default project
  • start – start position
  • length – length limit
  • compress_option (CompressOption) – the compression algorithm, level and strategy
Example:
>>> with odps.open_volume_reader('parted_volume', 'partition', 'file') as reader:
>>>     [print(line) for line in reader]
open_volume_writer(volume, partition=None, project=None, **kwargs)[source]

Write data into a volume. This function behaves differently under different types of volumes.

Under partitioned volumes, all files under a partition should be uploaded in one submission. The method returns a writer object with whose open method you can open a file inside the volume and write to it, or you can use write method to write to specific files.

Under file system volumes, the method returns a file-like object.

Parameters:
  • volume (str) – name of the volume
  • partition (str) – partition name for partitioned volumes, and path for file system volumes.
  • project (str) – project name, if not provided, will be the default project
  • compress_option (odps.tunnel.CompressOption) – the compression algorithm, level and strategy
Example:
>>> # Writing to partitioned volumes
>>> with odps.open_volume_writer('parted_volume', 'partition') as writer:
>>>     # both write methods are acceptable
>>>     writer.open('file1').write('some content')
>>>     writer.write('file2', 'some content')
>>> # Writing to file system volumes
>>> with odps.open_volume_writer('/fs_volume/dir1/file_name') as writer:
>>>     writer.write('some content')
read_table(name, limit=None, start=0, step=None, project=None, partition=None, **kw)[source]

Read table’s records.

Parameters:
  • name (odps.models.table.Table or str) – table or table name
  • limit – the records’ size, if None will read all records from the table
  • start – the record where read starts with
  • step – default as 1
  • project – project name, if not provided, will be the default project
  • partition – the partition of this table to read
  • columns (list) – the columns’ names which are the parts of table’s columns
  • compress (bool) – if True, the data will be compressed during downloading
  • compress_option (odps.tunnel.CompressOption) – the compression algorithm, level and strategy
  • endpoint – tunnel service URL
  • reopen – reading the table will reuse the session which opened last time, if set to True will open a new download session, default as False
Returns:

records

Return type:

generator

Example:
>>> for record in odps.read_table('test_table', 100):
>>>     # deal with such 100 records
>>> for record in odps.read_table('test_table', partition='pt=test', start=100, limit=100):
>>>     # read the `pt=test` partition, skip 100 records and read 100 records
run_archive_table(table, partition=None, project=None, hints=None, priority=None)[source]

Start running a task to archive tables.

Parameters:
  • table – name of the table to archive
  • partition – partition to archive
  • project – project name, if not provided, will be the default project
  • hints – settings for table archive task.
  • priority – instance priority, 9 as default
Returns:

instance

Return type:

odps.models.Instance

run_merge_files(table, partition=None, project=None, hints=None, priority=None)[source]

Start running a task to merge multiple files in tables.

Parameters:
  • table – name of the table to optimize
  • partition – partition to optimize
  • project – project name, if not provided, will be the default project
  • hints – settings for merge task.
  • priority – instance priority, 9 as default
Returns:

instance

Return type:

odps.models.Instance

run_security_query(query, project=None, token=None)[source]

Run a security query to grant / revoke / query privileges

Parameters:
  • query – query text
  • project – project name, if not provided, will be the default project
Returns:

a JSON object representing the result.

run_sql(sql, project=None, priority=None, running_cluster=None, hints=None, aliases=None, **kwargs)[source]

Run a given SQL statement asynchronously

Parameters:
  • sql (str) – SQL statement
  • project – project name, if not provided, will be the default project
  • priority (int) – instance priority, 9 as default
  • running_cluster – cluster to run this instance
  • hints (dict) – settings for SQL, e.g. odps.mapred.map.split.size
  • aliases (dict) –
Returns:

instance

Return type:

odps.models.Instance

run_xflow(xflow_name, xflow_project=None, parameters=None, project=None, hints=None, priority=None)[source]

Run xflow by given name, xflow project, paremeters asynchronously.

Parameters:
  • xflow_name (str) – XFlow name
  • xflow_project (str) – the project XFlow deploys
  • parameters (dict) – parameters
  • project – project name, if not provided, will be the default project
  • hints (dict) – execution hints
  • priority (int) – instance priority, 9 as default
Returns:

instance

Return type:

odps.models.Instance

set_project_policy(policy, project=None)[source]

Set policy of a project

Parameters:
  • policy – name of policy.
  • project – project name, if not provided, will be the default project
Returns:

JSON object

set_role_policy(name, policy, project=None)[source]

Get policy object of project

Parameters:
  • name – name of the role
  • policy – policy string or JSON object
  • project – project name, if not provided, will be the default project
set_security_option(option_name, value, project=None)[source]

Set a security option of a project

Parameters:
  • option_name – name of the security option. Please refer to ODPS options for more details.
  • value – value of security option to be set.
  • project – project name, if not provided, will be the default project.
stop_instance(id_, project=None)[source]

Stop the running instance by given instance id.

Parameters:
  • id – instance id
  • project – project name, if not provided, will be the default project
Returns:

None

stop_job(id_, project=None)

Stop the running instance by given instance id.

Parameters:
  • id – instance id
  • project – project name, if not provided, will be the default project
Returns:

None

write_table(name, *block_records, **kw)[source]

Write records into given table.

Parameters:
  • name (models.table.Table or str) – table or table name
  • block_records – if given records only, the block id will be 0 as default.
  • project – project name, if not provided, will be the default project
  • partition – the partition of this table to write
  • compress (bool) – if True, the data will be compressed during uploading
  • compress_option (odps.tunnel.CompressOption) – the compression algorithm, level and strategy
  • endpoint – tunnel service URL
  • reopen – writing the table will reuse the session which opened last time, if set to True will open a new upload session, default as False
Returns:

None

Example:
>>> odps.write_table('test_table', records)  # write to block 0 as default
>>>
>>> odps.write_table('test_table', 0, records)  # write to block 0 explicitly
>>>
>>> odps.write_table('test_table', 0, records1, 1, records2)  # write to multi-blocks
>>>
>>> odps.write_table('test_table', records, partition='pt=test') # write to certain partition
class odps.models.Project(**kw)[source]

Project is the counterpart of database in a RDBMS.

By get an object of Project, users can get the properties like name, owner, comment, creation_time, last_modified_time, and so on.

These properties will not load from remote ODPS service, unless users try to get them explicitly. If users want to check the newest status, try use reload method.

Example:
>>> project = odps.get_project('my_project')
>>> project.last_modified_time  # this property will be fetched from the remote ODPS service
>>> project.last_modified_time  # Once loaded, the property will not bring remote call
>>> project.owner  # so do the other properties, they are fetched together
>>> project.reload()  # force to update each properties
>>> project.last_modified_time  # already updated
class odps.models.Table(**kwargs)[source]

Table means the same to the RDBMS table, besides, a table can consist of partitions.

Table’s properties are the same to the ones of odps.models.Project, which will not load from remote ODPS service until users try to get them.

In order to write data into table, users should call the open_writer method with with statement. At the same time, the open_reader method is used to provide the ability to read records from a table or its partition.

Example:
>>> table = odps.get_table('my_table')
>>> table.owner  # first will load from remote
>>> table.reload()  # reload to update the properties
>>>
>>> for record in table.head(5):
>>>     # check the first 5 records
>>> for record in table.head(5, partition='pt=test', columns=['my_column'])
>>>     # only check the `my_column` column from certain partition of this table
>>>
>>> with table.open_reader() as reader:
>>>     count = reader.count  # How many records of a table or its partition
>>>     for record in record[0: count]:
>>>         # read all data, actually better to split into reading for many times
>>>
>>> with table.open_writer() as writer:
>>>     writer.write(records)
>>> with table.open_writer(partition='pt=test', blocks=[0, 1]):
>>>     writer.write(0, gen_records(block=0))
>>>     writer.write(1, gen_records(block=1))  # we can do this parallel
create_partition(partition_spec, if_not_exists=False, async_=False, **kw)[source]

Create a partition within the table.

Parameters:
  • partition_spec – specification of the partition.
  • if_not_exists
  • async
Returns:

partition object

Return type:

odps.models.partition.Partition

delete_partition(partition_spec, if_exists=False, async_=False, **kw)[source]

Delete a partition within the table.

Parameters:
  • partition_spec – specification of the partition.
  • if_exists
  • async
drop(async_=False, if_exists=False, **kw)[source]

Drop this table.

Parameters:async – run asynchronously if True
Returns:None
exist_partition(partition_spec)[source]

Check if a partition exists within the table.

Parameters:partition_spec – specification of the partition.
exist_partitions(prefix_spec=None)[source]

Check if partitions with provided conditions exist.

Parameters:prefix_spec – prefix of partition
Returns:whether partitions exist
get_ddl(with_comments=True, if_not_exists=False)[source]

Get DDL SQL statement for the given table.

Parameters:with_comments – append comment for table and each column
Returns:DDL statement
get_partition(partition_spec)[source]

Get a partition with given specifications.

Parameters:partition_spec – specification of the partition.
Returns:partition object
Return type:odps.models.partition.Partition
head(limit, partition=None, columns=None)[source]

Get the head records of a table or its partition.

Parameters:
  • limit – records’ size, 10000 at most
  • partition – partition of this table
  • columns (list) – the columns which is subset of the table columns
Returns:

records

Return type:

list

iterate_partitions(spec=None)[source]

Create an iterable object to iterate over partitions.

Parameters:spec – specification of the partition.
new_record(values=None)[source]

Generate a record of the table.

Parameters:values (list) – the values of this records
Returns:record
Return type:odps.models.Record
Example:
>>> table = odps.create_table('test_table', schema=Schema.from_lists(['name', 'id'], ['sring', 'string']))
>>> record = table.new_record()
>>> record[0] = 'my_name'
>>> record[1] = 'my_id'
>>> record = table.new_record(['my_name', 'my_id'])
open_reader(partition=None, **kw)[source]

Open the reader to read the entire records from this table or its partition.

Parameters:
  • partition – partition of this table
  • reopen (bool) – the reader will reuse last one, reopen is true means open a new reader.
  • endpoint – the tunnel service URL
  • compress_option (odps.tunnel.CompressOption) – compression algorithm, level and strategy
  • compress_algo – compression algorithm, work when compress_option is not provided, can be zlib, snappy
  • compress_level – used for zlib, work when compress_option is not provided
  • compress_strategy – used for zlib, work when compress_option is not provided
  • download_id – use existing download_id to download table contents
Returns:

reader, count means the full size, status means the tunnel status

Example:
>>> with table.open_reader() as reader:
>>>     count = reader.count  # How many records of a table or its partition
>>>     for record in reader[0: count]:
>>>         # read all data, actually better to split into reading for many times
open_writer(partition=None, blocks=None, **kw)[source]

Open the writer to write records into this table or its partition.

Parameters:
  • partition – partition of this table
  • blocks – block ids to open
  • reopen (bool) – the reader will reuse last one, reopen is true means open a new reader.
  • create_partition (bool) – if true, the partition will be created if not exist
  • endpoint – the tunnel service URL
  • upload_id – use existing upload_id to upload data
  • compress_option (odps.tunnel.CompressOption) – compression algorithm, level and strategy
  • compress_algo – compression algorithm, work when compress_option is not provided, can be zlib, snappy
  • compress_level – used for zlib, work when compress_option is not provided
  • compress_strategy – used for zlib, work when compress_option is not provided
Returns:

writer, status means the tunnel writer status

Example:
>>> with table.open_writer() as writer:
>>>     writer.write(records)
>>> with table.open_writer(partition='pt=test', blocks=[0, 1]):
>>>     writer.write(0, gen_records(block=0))
>>>     writer.write(1, gen_records(block=1))  # we can do this parallel
to_df()[source]

Create a PyODPS DataFrame from this table.

Returns:DataFrame object
truncate(async_=False, **kw)[source]

truncate this table.

Parameters:async – run asynchronously if True
Returns:None
odps.models.Schema

alias of odps.models.table.TableSchema

class odps.models.table.TableSchema(**kwargs)[source]

Schema includes the columns and partitions information of a odps.models.Table.

There are two ways to initialize a Schema object, first is to provide columns and partitions, the second way is to call the class method from_lists. See the examples below:

Example:
>>> columns = [Column(name='num', type='bigint', comment='the column')]
>>> partitions = [Partition(name='pt', type='string', comment='the partition')]
>>> schema = Schema(columns=columns, partitions=partitions)
>>> schema.columns
[<column num, type bigint>, <partition pt, type string>]
>>>
>>> schema = Schema.from_lists(['num'], ['bigint'], ['pt'], ['string'])
>>> schema.columns
[<column num, type bigint>, <partition pt, type string>]
class odps.models.partition.Partition(**kwargs)[source]

A partition is a collection of rows in a table whose partition columns are equal to specific values.

In order to write data into partition, users should call the open_writer method with with statement. At the same time, the open_reader method is used to provide the ability to read records from a partition. The behavior of these methods are the same as those in Table class except that there are no ‘partition’ params.

drop(async_=False, if_exists=False, **kw)[source]

Drop this partition.

Parameters:async – run asynchronously if True
Returns:None
open_reader(**kw)[source]

Open the reader to read the entire records from this partition.

Parameters:
  • reopen (bool) – the reader will reuse last one, reopen is true means open a new reader.
  • endpoint – the tunnel service URL
  • compress_option (odps.tunnel.CompressOption) – compression algorithm, level and strategy
  • compress_algo – compression algorithm, work when compress_option is not provided, can be zlib, snappy
  • compress_level – used for zlib, work when compress_option is not provided
  • compress_strategy – used for zlib, work when compress_option is not provided
Returns:

reader, count means the full size, status means the tunnel status

Example:
>>> with partition.open_reader() as reader:
>>>     count = reader.count  # How many records of a partition
>>>     for record in reader[0: count]:
>>>         # read all data, actually better to split into reading for many times
to_df()[source]

Create a PyODPS DataFrame from this partition.

Returns:DataFrame object
class odps.models.Record(columns=None, schema=None, values=None)[source]

A record generally means the data of a single line in a table.

Example:
>>> schema = Schema.from_lists(['name', 'id'], ['string', 'string'])
>>> record = Record(schema=schema, values=['test', 'test2'])
>>> record[0] = 'test'
>>> record[0]
>>> 'test'
>>> record['name']
>>> 'test'
>>> record[0:2]
>>> ('test', 'test2')
>>> record[0, 1]
>>> ('test', 'test2')
>>> record['name', 'id']
>>> for field in record:
>>>     print(field)
('name', u'test')
('id', u'test2')
>>> len(record)
2
>>> 'name' in record
True
class odps.models.Instance(**kwargs)[source]

Instance means that a ODPS task will sometimes run as an instance.

status can reflect the current situation of a instance. is_terminated method indicates if the instance has finished. is_successful method indicates if the instance runs successfully. wait_for_success method will block the main process until the instance has finished.

For a SQL instance, we can use open_reader to read the results.

Example:
>>> instance = odps.execute_sql('select * from dual')  # this sql return the structured data
>>> with instance.open_reader() as reader:
>>>     # handle the record
>>>
>>> instance = odps.execute_sql('desc dual')  # this sql do not return structured data
>>> with instance.open_reader() as reader:
>>>    print(reader.raw)  # just return the raw result
exception DownloadSessionCreationError(msg, request_id=None, code=None, host_id=None, instance_id=None)[source]
class Status[source]
class Task(**kwargs)[source]

Task stands for each task inside an instance.

It has a name, a task type, the start to end time, and a running status.

class TaskProgress(**kwargs)[source]

TaskProgress reprents for the progress of a task.

A single TaskProgress may consist of several stages.

Example:
>>> progress = instance.get_task_progress('task_name')
>>> progress.get_stage_progress_formatted_string()
2015-11-19 16:39:07 M1_Stg1_job0:0/0/1[0%]  R2_1_Stg1_job0:0/0/1[0%]
class TaskStatus[source]
class TaskSummary(*args, **kwargs)[source]
get_logview_address(hours=None)[source]

Get logview address of the instance object by hours.

Parameters:hours
Returns:logview address
Return type:str
get_sql_task_cost()[source]

Get cost information of the sql task. Including input data size, number of UDF, Complexity of the sql task

Returns:cost info in dict format
get_task_cost(task_name)[source]

Get task cost

Parameters:task_name – name of the task
Returns:task cost
Return type:Instance.TaskCost
Example:
>>> cost = instance.get_task_cost(instance.get_task_names()[0])
>>> cost.cpu_cost
200
>>> cost.memory_cost
4096
>>> cost.input_size
0
get_task_detail(task_name)[source]

Get task’s detail

Parameters:task_name – task name
Returns:the task’s detail
Return type:list or dict according to the JSON
get_task_detail2(task_name)[source]

Get task’s detail v2

Parameters:task_name – task name
Returns:the task’s detail
Return type:list or dict according to the JSON
get_task_info(task_name, key)[source]

Get task related information.

Parameters:
  • task_name – name of the task
  • key – key of the information item
Returns:

a string of the task information

get_task_names()[source]

Get names of all tasks

Returns:task names
Return type:list
get_task_progress(task_name)[source]

Get task’s current progress

Parameters:task_name – task_name
Returns:the task’s progress
Return type:odps.models.Instance.Task.TaskProgress
get_task_quota(task_name)[source]

Get queueing info of the task. Note that time between two calls should larger than 30 seconds, otherwise empty dict is returned.

Parameters:task_name – name of the task
Returns:quota info in dict format
get_task_result(task_name)[source]

Get a single task result.

Parameters:task_name – task name
Returns:task result
Return type:str
get_task_results()[source]

Get all the task results.

Returns:a dict which key is task name, and value is the task result as string
Return type:dict
get_task_statuses()[source]

Get all tasks’ statuses

Returns:a dict which key is the task name and value is the odps.models.Instance.Task object
Return type:dict
get_task_summary(task_name)[source]

Get a task’s summary, mostly used for MapReduce.

Parameters:task_name – task name
Returns:summary as a dict parsed from JSON
Return type:dict
get_task_workers(task_name=None, json_obj=None)[source]

Get workers from task :param task_name: task name :param json_obj: json object parsed from get_task_detail2 :return: list of workers

get_worker_log(log_id, log_type, size=0)[source]

Get logs from worker.

Parameters:
  • log_id – id of log, can be retrieved from details.
  • log_type – type of logs. Possible log types contains coreinfo, hs_err_log, jstack, pstack, stderr, stdout, waterfall_summary
  • size – length of the log to retrieve
Returns:

log content

is_running(retry=False)[source]

If this instance is still running.

Returns:True if still running else False
Return type:bool
is_successful(retry=False)[source]

If the instance runs successfully.

Returns:True if successful else False
Return type:bool
is_terminated(retry=False)[source]

If this instance has finished or not.

Returns:True if finished else False
Return type:bool
open_reader(*args, **kwargs)[source]

Open the reader to read records from the result of the instance. If tunnel is True, instance tunnel will be used. Otherwise conventional routine will be used. If instance tunnel is not available and tunnel is not specified,, the method will fall back to the conventional routine. Note that the number of records returned is limited unless options.limited_instance_tunnel is set to True or limit=True is configured under instance tunnel mode. Otherwise the number of records returned is always limited.

Parameters:
  • tunnel – if true, use instance tunnel to read from the instance. if false, use conventional routine. if absent, options.tunnel.use_instance_tunnel will be used and automatic fallback is enabled.
  • limit (bool) – if True, enable the limitation
  • reopen (bool) – the reader will reuse last one, reopen is true means open a new reader.
  • endpoint – the tunnel service URL
  • compress_option (odps.tunnel.CompressOption) – compression algorithm, level and strategy
  • compress_algo – compression algorithm, work when compress_option is not provided, can be zlib, snappy
  • compress_level – used for zlib, work when compress_option is not provided
  • compress_strategy – used for zlib, work when compress_option is not provided
Returns:

reader, count means the full size, status means the tunnel status

Example:
>>> with instance.open_reader() as reader:
>>>     count = reader.count  # How many records of a table or its partition
>>>     for record in reader[0: count]:
>>>         # read all data, actually better to split into reading for many times
put_task_info(task_name, key, value)[source]

Put information into a task.

Parameters:
  • task_name – name of the task
  • key – key of the information item
  • value – value of the information item
stop()[source]

Stop this instance.

Returns:None
wait_for_completion(interval=1)[source]

Wait for the instance to complete, and neglect the consequence.

Parameters:interval – time interval to check
Returns:None
wait_for_success(interval=1)[source]

Wait for instance to complete, and check if the instance is successful.

Parameters:interval – time interval to check
Returns:None
Raise:odps.errors.ODPSError if the instance failed
class odps.models.Resource(**kwargs)[source]

Resource is useful when writing UDF or MapReduce. This is an abstract class.

Basically, resource can be either a file resource or a table resource. File resource can be file, py, jar, archive in details.

class Type[source]
class odps.models.FileResource(**kw)[source]

File resource represents for a file.

Use open method to open this resource as an file-like object.

class Mode[source]
close()[source]

Close this file resource.

Returns:None
flush()[source]

Commit the change to ODPS if any change happens. Close will do this automatically.

Returns:None
open(mode='r', encoding='utf-8')[source]

The argument mode stands for the open mode for this file resource. It can be binary mode if the ‘b’ is inside. For instance, ‘rb’ means opening the resource as read binary mode while ‘r+b’ means opening the resource as read+write binary mode. This is most import when the file is actually binary such as tar or jpeg file, so be aware of opening this file as a correct mode.

Basically, the text mode can be ‘r’, ‘w’, ‘a’, ‘r+’, ‘w+’, ‘a+’ just like the builtin python open method.

  • r means read only
  • w means write only, the file will be truncated when opening
  • a means append only
  • r+ means read+write without constraint
  • w+ will truncate first then opening into read+write
  • a+ can read+write, however the written content can only be appended to the end
Parameters:
  • mode – the mode of opening file, described as above
  • encoding – utf-8 as default
Returns:

file-like object

Example:
>>> with resource.open('r') as fp:
>>>     fp.read(1)  # read one unicode character
>>>     fp.write('test')  # wrong, cannot write under read mode
>>>
>>> with resource.open('wb') as fp:
>>>     fp.readlines() # wrong, cannot read under write mode
>>>     fp.write('hello world') # write bytes
>>>
>>> with resource.open('test_resource', 'r+') as fp: # open as read-write mode
>>>     fp.seek(5)
>>>     fp.truncate()
>>>     fp.flush()
read(size=-1)[source]

Read the file resource, read all as default.

Parameters:size – unicode or byte length depends on text mode or binary mode.
Returns:unicode or bytes depends on text mode or binary mode
Return type:str or unicode(Py2), bytes or str(Py3)
readline(size=-1)[source]

Read a single line.

Parameters:size – If the size argument is present and non-negative, it is a maximum byte count (including the trailing newline) and an incomplete line may be returned. When size is not 0, an empty string is returned only when EOF is encountered immediately
Returns:unicode or bytes depends on text mode or binary mode
Return type:str or unicode(Py2), bytes or str(Py3)
readlines(sizehint=-1)[source]

Read as lines.

Parameters:sizehint – If the optional sizehint argument is present, instead of reading up to EOF, whole lines totalling approximately sizehint bytes (possibly after rounding up to an internal buffer size) are read.
Returns:lines
Return type:list
seek(pos, whence=0)[source]

Seek to some place.

Parameters:
  • pos – position to seek
  • whence – if set to 2, will seek to the end
Returns:

None

tell()[source]

Tell the current position

Returns:current position
truncate(size=None)[source]

Truncate the file resource’s size.

Parameters:size – If the optional size argument is present, the file is truncated to (at most) that size. The size defaults to the current position.
Returns:None
write(content)[source]

Write content into the file resource

Parameters:content – content to write
Returns:None
writelines(seq)[source]

Write lines into the file resource.

Parameters:seq – lines
Returns:None
class odps.models.PyResource(**kw)[source]

File resource representing for the .py file.

class odps.models.JarResource(**kw)[source]

File resource representing for the .jar file.

class odps.models.ArchiveResource(**kw)[source]

File resource representing for the compressed file like .zip/.tgz/.tar.gz/.tar/jar

class odps.models.TableResource(**kw)[source]

Take a table as a resource.

open_reader(**kwargs)[source]

Open reader on the table resource

open_writer(**kwargs)[source]

Open writer on the table resource

partition

Get the source table partition.

Returns:the source table partition
table

Get the table object.

Returns:source table
Return type:odps.models.Table
update(project_name=None, table_name=None, *args, **kw)[source]

Update this resource.

Parameters:
  • project_name – the source table’s project
  • table_name – the source table’s name
  • partition – the source table’s partition
Returns:

self

class odps.models.Function(**kwargs)[source]

Function can be used in UDF when user writes a SQL.

drop()[source]

Delete this Function.

Returns:None
resources

Return all the resources which this function refer to.

Returns:resources
Return type:list
update()[source]

Update this function.

Returns:None
class odps.models.Worker(**kwargs)[source]

Worker information class for worker information and log retrieval.

get_log(log_type, size=0)[source]

Get logs from worker.

Parameters:
  • log_type – type of logs. Possible log types contains coreinfo, hs_err_log, jstack, pstack, stderr, stdout, waterfall_summary
  • size – length of the log to retrieve
Returns:

log content

class odps.models.ml.OfflineModel(**kwargs)[source]

Representing an ODPS offline model.

copy(new_name, new_project=None, async_=False)[source]

Copy current model into a new location.

Parameters:
  • new_name – name of the new model
  • new_project – new project name. if absent, original project name will be used
  • async – if True, return the copy instance. otherwise return the newly-copied model
get_model()[source]

Get PMML text of the current model. Note that model file obtained via this method might be incomplete due to size limitations.

class odps.models.security.User(**kwargs)[source]