Introduction¶
populse_db is a small Python module aiming is to store and query efficiently JSON data with the following constraints:
Storage on file system: The data must be stored on a standard file system and respect all system access rights.
Parallel access: Multiple processes or threads must be able to read or modify data in parallel.
Use a few files: populse_db store data in a single file. In some environment where the number of inodes is limited, it is vital that data storage have also a reasonable limit on the number of files it produces.
No schema required: populse_db can be used without thinking about database schema. But it is also possible to define some pieces of schema in order to optimize query performances on some JSON attributes or to define a partitioning of the data.
Date and time support: populse_db support items of type datetime.time, datetime.date and datetime.datetime in JSON objects. These items are automatically encoded (resp. decoded) to (resp. from) strings using ISO format.
Query list elements: populse_db has a simple query language that allows to select objects by looking for items in list fields.
Transactions support: to ensure the consistency of the database, populse_db allows all modifications to be made within a transaction. Thus, in case of unexpected problem, the database will keep its initial state before the transaction.
Partitioning of JSON objects: populse_db allows to read or modify individually each stored object. But it also allows to define a partitioning of the objects offering a direct access to different parts of the same object without having to read or write the whole object (as it would be using a simple JSON file).
populse_db offers a simple API on top of a SQLite backend that allows to respect all the requirements.
Dependencies¶
Install tools take care of all the required dependencies for you but here is a summary of what populse_db needs:
Python >= 3.9. populse_db uses some features introduced in Python 3.9 (such as the possibility to subscript list type as in
list[str]). Therefore it can’t work with Python 3.8 or earlier releases.Lark-parser >= 0.7.0.
[optional] Sphinx allows to build the documentation from source code.
Get started¶
Installation¶
Populse_db can be installed by standard Python tool. For instance:
pip install populse_db
The documentation is embedded in the project. However, if one needs to rebuild the documentation, use:
pip install populse_db[doc]
Basic usage¶
Populse_db is organized in collections and documents. A document is a JSON object represented by a dictionary in Python. A collection is a named container in which documents can be stored. Internally, collections corresponds to database tables and documents are stored in table rows (one document per row). But populse_db can hide this internal database stuff. The only thing user must do isto declare its collections if it uses an empty database.
from datetime import date, datetime
from populse_db import Storage
from pprint import pprint
# Create a storage using a file name
storage = Storage('/tmp/populse_db.sqlite')
# Create a read/write session
with storage.session(write=True) as db:
db.last_modified = datetime.now()
# Store documents in the database
db['subject'] = Storage.Collection('subject_id')
db['subject']['rbndt001'] = {
'name': 'Eléa',
'sex': 'f',
'birth_date': date(1968, 3, 3),
}
db['subject']['rbndt002'] = {
'name': 'Païkan',
'sex': 'm',
'birth_date': date(1963, 12, 7),
}
db['acquisition'] = Storage.Collection()
db['acquisition']['rbndt001_t1'] = {
'subject_id': 'rbndt001',
'type': ['image', 'mri', 'T1'],
'format': 'DICOM',
'files': [
'/somewhere/t1/acq0001.dcm',
'/somewhere/t1/acq0002.dcm',
'/somewhere/t1/acq0003.dcm',
'/somewhere/t1/acq0004.dcm',
],
'date': date(2022, 3, 28),
}
db['acquisition']['rbndt001_t2'] = {
'subject_id': 'rbndt001',
'type': ['image', 'mri', 'T2'],
'format': 'DICOM',
'files': [
'/somewhere/t2/acq0001.dcm',
'/somewhere/t2/acq0002.dcm',
'/somewhere/t2/acq0003.dcm',
'/somewhere/t2/acq0004.dcm',
],
'date': date(2022, 3, 28),
}
db['acquisition']['rbndt002_t1'] = {
'subject_id': 'rbndt002',
'type': ['image', 'mri', 'T1'],
'format': 'NIFTI',
'files': [
'/elsewhere/sub-rbndt001.nii',
],
'date': date(2022, 3, 29),
}
# Retrieve a single value from storage
print('Last modified:', db.last_modified.get())
# Retrieve a single document from collection 'subject'
pprint(db['subject']['rbndt001'].get())
# Retrieve all documents from collection 'subject' respecting the
# following conditions:
# - the "subject" field equals to "rbndt001"
# - the "type" field is a list containing the value "T1"
for doc in db['acquisition'].search('subject=="rbndt001" and "T1" in type'):
pprint(doc)
# Retrieve all possible types of acquisition
print('Subjects:', ', '.join(db.subject.distinct_values('subject_id')))
The example above illustrate how to use populse_db to store and retrieve JSON objects without wondering about the underlying SQL database engine. This script produces the following result:
{'birth_date': datetime.date(1968, 3, 3),
'name': 'Eléa',
'primary_key': 'rbndt001',
'sex': 'f'}
{'date': datetime.date(2022, 3, 28),
'files': ['/somewhere/t1/acq0001.dcm',
'/somewhere/t1/acq0002.dcm',
'/somewhere/t1/acq0003.dcm',
'/somewhere/t1/acq0004.dcm'],
'format': 'DICOM',
'primary_key': 'rbndt001_t1',
'subject': 'rbndt001',
'type': ['image', 'mri', 'T1']}