Data description and validation for CKAN with Frictionless Data tools.
This extension brings data validation powered by the Frictionless Framework library to CKAN. It provides out of the box features to validate tabular data and integrate validation reports to the CKAN interface.
Data validation can be performed automatically on the background or during dataset creation, and the results are stored against each resource.
Comprehensive reports are created describing issues found with the data, both at the structure level (missing headers, blank rows, etc) and at the data schema level (wrong data types, values out of range, etc).
The extension also exposes all the underlying actions so data validation can be integrated in custom workflows from other extensions.
If you are eager to get started, jump to the Installation and Configuration instructions. To learn more about data validation and how the extension works, read the next section.
This extension is currently tested in CKAN 2.9 (Python 3). For previous CKAN versions and Python 2 you can use the 1.x versions, but those will not be supported going forward.
It is strongly recommended to use it alongside ckanext-scheming to define the necessary extra fields in the default CKAN schema.
To install ckanext-validation, activate your CKAN virtualenv and run:
git clone https://github.com/frictionlessdata/ckanext-validation.git
cd ckanext-validation
pip install -r requirements.txt
python setup.py develop
Create the database tables running:
paster validation init-db -c ../path/to/ini/file
Once installed, add the validation
plugin to the ckan.plugins
configuration option in your INI file. If using ckanext-scheming, the validation
plugins should be loaded before the scheming_datasets
one:
ckan.plugins = ... validation scheming_datasets
The extension requires changes in the CKAN metadata schema. The easiest way to add those is by using ckanext-scheming. Use these two configuration options to link to the dataset schema (replace with your own if you need to customize it) and the required presets:
scheming.dataset_schemas = ckanext.validation.examples:ckan_default_schema.json
scheming.presets = ckanext.scheming:presets.json
ckanext.validation:presets.json
Read more below about how to change the CKAN metadata schema
Use the following configuration options to choose the operation modes:
ckanext.validation.run_on_create_async = True|False (Defaults to True)
ckanext.validation.run_on_update_async = True|False (Defaults to True)
ckanext.validation.run_on_create_sync = True|False (Defaults to False)
ckanext.validation.run_on_update_sync = True|False (Defaults to False)
By default validation will be run against the following formats: CSV
, XLSX
and XLS
. You can modify these formats using the following option:
ckanext.validation.formats = csv xlsx
You can also provide validation options that will be used by default when running the validation:
ckanext.validation.default_validation_options={
"skip_errors": ["blank-row", "duplicate-label"],
}
Make sure to use indentation if the value spans multiple lines otherwise it won't be parsed.
If you are using a cloud-based storage backend for uploads, check Private datasets for other configuration settings that might be relevant.
To prevent the extension from adding the validation badges next to the resources use the following option:
ckanext.validation.show_badges_in_listings = False
CKAN users will be familiar with the validation performed against the metadata fields when creating or updating datasets. The form will return an error, for instance, if a field is missing or it doesn't have the expected format.
Data validation follows the same principle, but against the actual data published in CKAN, that is the contents of tabular files (Excel, CSV, etc) hosted in CKAN itself or elsewhere. Whenever a resource of the appropriate format is created or updated, the extension will validate the data against a collection of checks. This validation is powered by Frictionless Framework, a very powerful data validation library developed by the Open Knowledge Foundation as part of the Frictionless Data project. Frictionless Framework provides an extensive suite of checks that cover common issues with tabular data files.
These checks include structural problems like missing headers or values, blank rows, etc., but also can validate the data contents themselves (see Data Schemas) or even run custom checks.
The result of this validation is a JSON report. This report contains all the issues found (if any) with their relevant context (row number, columns involved, etc). The reports are stored in the database and linked to the CKAN resources, and can be retrieved via the API.
If there is a report available for a particular resource, a status badge will be displayed in the resource listing and on the resource page, showing whether validation passed or failed for the resource.
Clicking on the badge will take you to the validation report page, where the report will be rendered.
Whenever possible, the report will provide a preview of the cells, rows or columns involved in an error, to make it easy to identify and fix it.
As mentioned before, data can be validated against a schema. Much in the same way as the standard CKAN schema for metadata fields, the schema describes the data and what its values are expected to be.
These schemas are defined following the Table Schema specification, a really simple and flexible standard for describing tabular data.
Let's see an example. Consider the following table (that could be stored as a CSV or Excel file):
id | location | date | measurement | observations |
---|---|---|---|---|
1 | 'A' | 01/02/2017 | 23.65 | |
2 | 'B' | 21/03/2017 | 22.90 | |
3 | 'A' | 15/06/2017 | 21.79 | Severe drought |
4 | 'C' | 10/10/2017 | 24.12 | |
5 | 'C' | 31/10/2017 | 24.21 |
The following schema describes the expected data:
{
"primaryKey": "id",
"fields": [
{
"name": "id",
"title": "Measurement identifier",
"type": "integer"
},
{
"name": "location",
"title": "Measurement location code",
"type": "string",
"constraints": {
"enum": ["A", "B", "C", "D"]
}
},
{
"name": "date",
"title": "Measurement date",
"type": "date",
"format": "%d/%m/%Y"
},
{
"name": "measurement",
"title": "Measure of the oblique fractal impedance at noon",
"type": "number",
"constraints": {
"required": true
}
},
{
"name": "observations",
"title": "Extra observations",
"type": "string"
}
]
}
If we store this schema against a resource, it will be used to perform a more thorough validation. For instance, updating the resource with the following data would fail validation with a variety of errors, even if the general structure of the file is correct:
id | location | date | measurement | observations |
---|---|---|---|---|
... | ... | ... | ... | ... |
5 | 'E' | 2017-11-01 | missing | |
'a' | 'B' | 21/03/2017 |
With the extension enabled and configured, schemas can be attached to the
schema
field on resources via the UI form or the API. If present in a
resource, they will be used when performing validation on the resource file.
As we saw before, the validation process involves many different checks and it's very likely that what "valid" data actually means will vary across CKAN instances or datasets. The validation process can be tweaked by passing any of the supported options to Frictionless Framework. These can be used to add or remove specific checks, control limits, etc.
For instance, the following file would fail validation using the default options, but it may be valid in a given context, or the issues may be known to the publishers:
<blank line>
<blank line>
id;group;measurement
# 2017
1;A;23
2;B;24
# 2016
3;C;23
4;C;25
<blank line>
The following validation options would make validation pass:
{
"skip_errors": ["blank-row"]
"dialect": {
"header": True,
"headerRows": [2],
"commentChar": "#",
"csv": {
"delimiter": ";"
}
},
"checks": [
{"type": "table-dimensions", "minRows": 3},
{"type": 'ascii-value'}
]
}
Validation options can be defined (as a JSON object like the above) on each
resource (via the UI form or the API on the validation_options
field) or can
be set globally by administrators on the CKAN INI file (see Configuration).
Validation can be performed on private datasets. When validating a locally
uploaded resource, the extension uses the actual physical path to read the
file, so internally there is no need for authorization. But when the upload is
on a cloud-based backend (like the ones provided by ckanext-cloudstorage or
ckanext-s3filestore) we need
to request the file via an HTTP request to CKAN. If the resource is private
this will require an Authorization
header in order to avoid a Not Authorized
error.
In these cases, the API key for the site user will be passed as part of the
request (or alternatively ckanext.validation.pass_auth_header_value
if set in
the configuration).
As this involves sending API keys to other extensions, this behaviour can be
turned off by setting ckanext.validation.pass_auth_header
to False
.
Again, these settings only affect private resources when using a cloud-based backend.
The data validation process described above can be run in two modes: asynchronously in the background or synchronously while the resource is being created or updated. You can choose the mode for each of the create and update actions, but in most cases you will probably need just one of the two modes for both actions.
Asynchronous validation is run in the background whenever a resource of a supported format is created or updated. Validation won't affect the action performed, so if there are validation errors found the resource will be created or updated anyway.
This mode might be useful for instances where datasets are harvested from other sources, or where multiple publishers create datasets and as a maintainer you only want to give visibility to the quality of data, encouraging publishers to fix any issues.
You will need to run the worker
commmand to pick up validation jobs. Please
refer to the background jobs documentation
for more details:
paster jobs worker -c /path/to/ini/file
Use ckanext.validation.run_on_create_async
and
ckanext.validation.run_on_update_async
to enable this mode (See Configuration).
Synchronous validation is performed at the same time a resource of the
supported formats is being created or updated. Currently, if data validation
errors are found, a ValidationError
will be raised and you won't be able to
create or update the resource.
Validation at creation or update time can be useful to ensure that data quality is maintained or that published data conforms to a particular schema.
When using the UI form, validation errors will be displayed as normal CKAN validation errors:
Clicking the link on the error message will bring up a modal window with the validation report rendered:
Use ckanext.validation.run_on_create_sync
and ckanext.validation.run_on_update_sync
to enable this mode (See Configuration).
The extension requires changes in the default CKAN resource metadata schema to
add some fields it requires. It is strongly recommended to use
ckanext-scheming to define your
CKAN schema. This extension provides all the necessary presets and validators
to get up and running just by adding the following fields to the
resource_fields
section of a ckanext-scheming schema:
{
"field_name": "schema",
"label": "Schema",
"preset": "resource_schema"
},
{
"field_name": "validation_options",
"label": "Validation options",
"preset": "validation_options"
},
{
"field_name": "validation_status",
"label": "Validation status",
"preset": "hidden_in_form"
},
{
"field_name": "validation_timestamp",
"label": "Validation timestamp",
"preset": "hidden_in_form"
}
Here's more detail on the fields added:
schema
: This can be a Table Schema
JSON object or an URL pointing to one. In the UI form you can upload a JSON file, link to one
providing a URL or enter it directly. If uploaded, the file contents will be
read and stored in the schema
field. In all three cases the contents will be
validated against the Table Schema specification.validation_options
: A JSON object with validation options that will be
passed to Frictionless Framework validate
function.Additionally, two read-only fields are added to resources:
validation_status
: Stores the last validation result for the resource.
Can be one of success
, failure
or error
.validation_timestamp
: Date and time of the last validation run.The plugin provides the IDataValidation
interface so other plugins can modify
its behaviour.
Currently it only provides the can_validate()
method, that plugins can use to
determine if a specific resource should be validated or not:
class IDataValidation(Interface):
def can_validate(self, context, data_dict):
'''
When implemented, this call can be used to control whether the
data validation should take place or not on a specific resource.
Implementations will receive a context object and the data_dict of
the resource.
If it returns False, the validation won't be performed, and if it
returns True there will be a validation job started.
Note that after this methods is called there are further checks
performed to ensure the resource has one of the supported formats.
This is controlled via the `ckanext.validation.formats` config option.
Here is an example implementation:
from ckan import plugins as p
from ckanext.validation.interfaces import IDataValidation
class MyPlugin(p.SingletonPlugin):
p.implements(IDataValidation, inherit=True)
def can_validate(self, context, data_dict):
if data_dict.get('my_custom_field') == 'xx':
return False
return True
'''
return True
The validation
plugin adds new API actions to create and display validation
reports. By default resource_validation_run
, resource_validation_delete
and
resource_validation_show
inherit whatever auth is in place for
resource_update
and resource_show
respectively.
There is an extra action which only sysadmins can access:
resource_validation_run_batch
.
resource_validation_run
def resource_validation_run(context, data_dict):
u'''
Start a validation job against a resource.
Returns the identifier for the job started.
Note that the resource format must be one of the supported ones,
currently CSV or Excel.
:param resource_id: id of the resource to validate
:type resource_id: string
:rtype: string
'''
resource_validation_show
def resource_validation_show(context, data_dict):
u'''
Display the validation job result for a particular resource.
Returns a validation object, including the validation report or errors
and metadata about the validation like the timestamp and current status.
Validation status can be one of:
* `created`: The validation job is in the processing queue
* `running`: Validation is under way
* `error`: There was an error while performing the validation, eg the file
could not be downloaded or there was an error reading it
* `success`: Validation was performed, and no issues were found
* `failure`: Validation was performed, and there were issues found
:param resource_id: id of the resource to validate
:type resource_id: string
:rtype: dict
'''
resource_validation_delete
def resource_validation_delete(context, data_dict):
u'''
Remove the validation job result for a particular resource.
It also deletes the underlying Validation object.
:param resource_id: id of the resource to remove validation from
:type resource_id: string
:rtype: None
'''
resource_validation_run_batch
def resource_validation_run_batch(context, data_dict):
u'''
Start asynchronous data validation on the site resources. If no
options are provided it will run validation on all resources of
the supported formats (`ckanext.validation.formats`). You can
specify particular datasets to run the validation on their
resources. You can also pass arbitrary search parameters to filter
the selected datasets.
Only sysadmins are allowed to run this action.
Examples::
curl -X POST http://localhost:5001/api/action/resource_validation_run_batch \
-d '{"dataset_ids": "ec9bfd88-f90a-45ca-b024-adc8854b49bd"}' \
-H Content-type:application/json \
-H Authorization:API_KEY
curl -X POST http://localhost:5001/api/action/resource_validation_run_batch \
-d '{"dataset_ids": ["passenger-data-2018", "passenger-data-2017]}}' \
-H Content-type:application/json \
-H Authorization:API_KEY
curl -X POST http://localhost:5001/api/action/resource_validation_run_batch \
-d '{"query": {"fq": "res_format:XLSX"}}' \
-H Content-type:application/json \
-H Authorization:API_KEY
:param dataset_ids: Run data validation on all resources for a
particular dataset or datasets. Not to be used with ``query``.
:type dataset_ids: string or list
:param query: Extra search parameters that will be used for getting
the datasets to run validation on. It must be a JSON object like
the one used by the `package_search` API call. Supported fields
are ``q``, ``fq`` and ``fq_list``. Check the documentation for
examples. Note that when using this you will have to specify
the resource formats to target your Not to be used with
``dataset_ids``.
:type query: dict
:rtype: string
'''
You can start (asynchronous) validation jobs from the command line using the
paster validation run
command. If no parameters are provided it will start a
validation job for all resources in the site of suitable format (ie
ckanext.validation.formats
):
paster validation run -c /path/to/ckan/ini
You can limit the resources by specifying a dataset id or name:
paster validation run -c /path/to/ckan/ini -d statistical-data-2018
Or providing arbitrary search parameters:
paster validation run -c ../ckan/development.ini -s '{"fq":"res_format:XLSX"}'
The extension provides two small utilities to generate a global report with all the current data validation reports:
paster validation report -c /path/to/ckan/ini
paster validation report-full -c /path/to/ckan/ini
Both commands will print an overview of the total number of datasets and
tabular resources, and a breakdown of how many have a validation status of
success, failure or error. Additionally they will create a CSV report. paster validation report
will create a report with all failing resources, including
the following fields:
paster validation report-full
will add a row on the output CSV for each error
found on the validation report (limited to ten occurrences of the same error
type per file). So the fields in the generated CSV report will be:
In both cases you can define the location of the output CSV passing the -o
or
--output
option:
paster validation report-full -c /path/to/ckan/ini -o /tmp/reports/validation_full.csv
Check the command help for more details:
paster validation --help
Usage: paster validation [options] Utilities for the CKAN data validation extension
Usage:
paster validation init-db
Initialize database tables
paster validation run [options]
Start asynchronous data validation on the site resources. If no
options are provided it will run validation on all resources of
the supported formats (`ckanext.validation.formats`). You can
specify particular datasets to run the validation on their
resources. You can also pass arbitrary search parameters to filter
the selected datasets.
paster validation report [options]
Generate a report with all current data validation reports. This
will print an overview of the total number of tabular resources
and a breakdown of how many have a validation status of success,
failure or error. Additionally it will create a CSV report with all
failing resources, including the following fields:
* Dataset name
* Resource id
* Resource URL
* Status
* Validation report URL
paster validation report-full [options]
Generate a detailed report. This is similar to the previous command
but on the CSV report it will add a row for each error found on the
validation report (limited to ten occurrences of the same error
type per file). So the fields in the generated CSV report will be:
* Dataset name
* Resource id
* Resource URL
* Status
* Error code
* Error message
Options:
-h, --help show this help message and exit
-v, --verbose
-c CONFIG, --config=CONFIG
Config file to use.
-f FILE_PATH, --file=FILE_PATH
File to dump results to (if needed)
-y, --yes Automatic yes to prompts. Assume "yes" as answer to
all prompts and run non-interactively
-r RESOURCE_ID, --resource=RESOURCE_ID
Run data validation on a particular resource (if the
format is suitable). It can be defined multiple times.
Not to be used with -d or -s
-d DATASET_ID, --dataset=DATASET_ID
Run data validation on all resources for a particular
dataset (if the format is suitable). You can use the
dataset id or name, and it can be defined multiple
times. Not to be used with -r or -s
-s SEARCH_PARAMS, --search=SEARCH_PARAMS
Extra search parameters that will be used for getting
the datasets to run validation on. It must be a JSON
object like the one used by the `package_search` API
call. Supported fields are `q`, `fq` and `fq_list`.
Check the documentation for examples. Note that when
using this you will have to specify the resource
formats to target yourself. Not to be used with -r or
-d.
-o OUTPUT_FILE, --output=OUTPUT_FILE
Location of the CSV validation report file on the
relevant commands.
To run the tests, do:
pip install -r dev-requirements.txt
pytest --ckan-ini=test-custom.ini
This material is copyright (c) Open Knowledge Foundation.
It is open and licensed under the GNU Affero General Public License (AGPL) v3.0 whose full text may be found at: