How to…#

On other pages:

Access other SDMX data sources#

sdmx ships with a file, sources.json, that includes information about the capabilities of many data sources. However, any data source that generates SDMX 2.1 messages is supported. There are multiple ways to access these:

  1. Create a sdmx.Client without a named data source, then call the get() method using the url argument:

    import sdmx
    c = sdmx.Client()
    c.get(url='', ...)
  2. Call add_source() with a JSON snippet describing the data provider.

  3. Create a subclass of Source, providing attribute values and optional implementations of hooks.

Control logging#

sdmx.log is a standard Python logging.Logger object. For debugging, set this to a permissive level:

import logging


Log messages include the web service query details (URL and headers) used by Client.

Use the ‘references’ query parameter#

SDMX web services support a references parameter in HTTP requests which can take values such as ‘all’, ‘descendants’, etc. This parameter instructs the web service to include, when generating a Data- or StructureMessage, the objects implicitly designated by the references parameter alongside the explicit resource. For example, for the request:

>>> response = some_agency.dataflow('SOME_ID', params={'references': 'all'})

the response will include:

  • the dataflow ‘SOME_ID’ explicitly specified,

  • the DSD referenced by the dataflow’s structure attribute,

  • the code lists referenced by the DSD, and

  • any content-constraints which reference the dataflow or the DSD.

It is much more efficient to request many objects in a single request. Thus, sdmx provides default values for references in common queries. For instance, when a single dataflow is requested by specifying its ID, sdmx sets references to ‘all’. On the other hand, when the dataflow ID is wild-carded, it is more practical not to request all referenced objects alongside as the response would likely be excessively large, and the user is deemed to be interested in the bird’s eye perspective (list of data flows) prior to focusing on a particular dataflow and its descendants and ancestors. The default value for the references parameter can be overridden.

Some web services differ in how they handle references—for instance, ESTAT. See Data sources for details.

Use category schemes to explore data#

SDMX supports category-schemes to categorize dataflow definitions and other objects. This helps retrieve, e.g., a dataflow of interest. Note that not all agencies support category schemes. A good example is the ECB. However, as the ECB’s SDMX service offers less than 100 data flows, using category schemes is not strictly necessary. A counter-example is Eurostat which offers more than 6000 data flows, yet does not categorize them. Hence, the user must search through the flat list of data flows.

To search the list of data flows by category, we request the category scheme from the ECB’s SDMX service and explore the response:

In [1]: import sdmx

In [2]: ecb = sdmx.Client('ecb')

In [3]: cat_response = ecb.categoryscheme()

Like any other scheme, a category scheme is essentially a dict mapping ID’s to the actual SDMX objects. To display the categorized items, in our case the dataflow definitions contained in the category on exchange rates, we iterate over the Category instance:

In [4]: sdmx.to_pandas(cat_response.category_scheme.MOBILE_NAVI)
01                                  Monetary operations
02             Prices, output, demand and labour market
03                    Monetary and financial statistics
04                                   Euro area accounts
05                                   Government finance
06                  External transactions and positions
07                                       Exchange rates
08    Payments and securities trading, clearing, set...
09                                  Banknotes and Coins
10                  Indicators of Financial Integration
11               Real Time Database (research database)
Name: Economic concepts, dtype: object

In [5]: cat_response.category_scheme.MOBILE_NAVI
Out[5]: <CategoryScheme ECB:MOBILE_NAVI(1.0) (11 items): Economic concepts>

Added in version 0.5.

Select data frame layouts returned by to_pandas()#

to_pandas() provides multiple ways to customize the type and layout of pandas objects returned for DataMessage input. One is the datetime argument; see Convert dimensions to pandas.DatetimeIndex or PeriodIndex. The other is the rtype argument.

To select the same behaviour as pandaSDMX 0.9, give rtype = ‘compat’, or set DEFAULT_RTYPE to ‘compat’:

In [6]: sdmx.writer.DEFAULT_RYPE = 'compat'

With ‘compat’, the returned layout varies with the concept of “dimension at the observation level,” as follows:

Dimension At Observation Level

Return Type



Same as datetime = True —a Dataframe with:

Other Dimension

DataFrame with:

  • index: the dimension at observation level, and

  • columns: MultiIndex with all other dimensions.


  • sdmx can only obey rtype = ‘compat’ when reading or converting an entire DataMessage; not a DataSet. While the concept of “dimension at observation level” is mentioned in the IM in relation to data sets, it is not formally included as an attribute of any class, or with any default value. (For instance, it is not included in the DimensionDescriptor of a DataStructureDefinition.) It can only be determined from the header of a SDMX-ML or -JSON data message.

  • Except for AllDimensions, each row and column of the returned data frame contains multiple observations, so attributes cannot be included without ambiguity about which observation(s) have the attribute. In these cases, attributes are omitted; use rtype = ‘rows’ to retrieve them.

With the argument rtype = ‘rows’ (the default), data are always returned with one row per observation.

Convert SDMX data to other formats#

Pandas supports output to many popular file formats. Call these methods on the objects returned by to_pandas(). For instance:

msg = sdmx.read_sdmx('data.xml')

pandaSDMX 0.9 could be used with odo by registering methods for discovery and conversion:

import odo
from odo.utils import keywords
import pandas as pd
from toolz import keyfilter
import toolz.curried.operator as op

class PandaSDMX(object):
    def __init__(self, uri):
        self.uri = uri

def _resource(uri, **kwargs):
    return PandaSDMX(uri)
def _discover(obj):

@odo.convert.register(pd.DataFrame, PandaSDMX)
def _convert(obj, **kwargs):
    msg = sdmx.read_sdmx(obj.uri)
    return sdmx.to_pandas(
        msg, **keyfilter(op.contains(keywords(write)), kwargs)

Deprecated since version 1.0: odo appears unmaintained since about 2016, so sdmx no longer provides built-in registration.

Added in version 0.4: sdmx.odo_register() was added, providing automatic registration.