Skip to content

Latest commit

 

History

History
274 lines (199 loc) · 13.3 KB

README.md

File metadata and controls

274 lines (199 loc) · 13.3 KB

utilix

PyPI version shields.io pre-commit.ci status

utilix is a utility package for XENON software, mainly relating to analysis. It currently has two main features: (1) a general XENON configuration framework and (2) easy access to the runsDB by wrapping python calls to a RESTful API. Eventually, we would like to include easy functions for interacting with the Midway and OSG batch queues.

Installation

git clone this repo and:

# at top level of utilix
pip install -e ./ --user

Note: you may need to add --ignore-installed at the end of pip install if it tries to uninstall the old package from a public path (for example, under /cvmfs/xenon.opensciencegrid.org/releases/nT/2024.02.1/anaconda/envs/XENONnT_2024.02.1/lib/python3.9/site-packages) which you don't have access to.

Configuration file

This tool expects a configuration file given by the environment variable XENON_CONFIG, defaulting to $HOME/.xenon_config if it is empty. Note that environment variables can be used in the form $HOME. Example:

[RunDB]
rundb_api_url = [ask teamA]
rundb_api_user = [ask teamA]
rundb_api_password = [ask teamA]

The idea is that analysts could use this single config for multiple purposes/analyses. You just need to add a (unique) section for your own purpose and then you can use the utilix.Config easily. For example, if you made a new section called WIMP with detected = yes under it:

from utilix.config import Config
cfg = Config()
value = cfg.get('WIMP', 'detected') # value = 'yes'

For more information, see the ConfigParser documentation, from which utilix.config.Config inherits.

Runs Database

Nearly every analysis requires access to the XENON runsDB. The goal of utilix is to simplify the usage of this resource as much as possible. The rundb module includes two ways to access the runsDB:

  1. A RESTful API: a Flask app running at Chicago that queries the runDB in a controlled manner. This is the recommended way to query the database if the specific query is supported. The source code for this app can be found here.
  2. A wrapper around pymongo, which sets up the Mongo client for you, similarly to how we did queries in XENON1T. In that case each package usually needed its own copy + pasted boilerplate code; that code is now just included in utilix where it can be easily imported by other packages.

RunDB API

RunDB API Authentication

The API authenticates using a token system. utilix makes the creation and renewal of these tokens easy with the utilix.rundb.Token class. When you specify a user/password in your utilix configuration file, as shown above, a token is saved locally at ~/.dbtoken that contains this information. This token is used/renewed as needed, depending on the users specified in the config file.

Different API users have different permissions, with the general analysis user only able to read from the runDB and not write. This is an additional layer of security around the RunDB.

Setting up the runDB

The goal of utilix is to make access to the runDB trivial. If using the runDB API, all you need to do to setup the runDB in your local script/shell is

from utilix import db

This instantiates the RunDB class, allowing for easy queries. Below we go through some examples of the type of queries currently supported by the runDB API wrapper in utilix.

If there is functionality missing that you think would be useful, please contact teamA or make a new issue (or even better, a pull request).

Query for runs by source

Note that the interface returns pages of 1,000 entries, with the first page being 1.

from utilix import db
data = db.query_by_source('neutron_generator', page_num=1)

Get a full document

You can also grab the full run document using the run number. A run name is also supported (from XENON1T days), but not going to be used for XENONnT

doc = db.get_doc(7200)

Get only the data entry of a document

data = db.get_data(2000)

Strax(en) Contexts

In XENONnT we need to track the hash (or lineage) that specifies a configuration for each datatype. We keep that information in a specific collection of the runDB. We can access that collection using the runDB API as shown below.

For a given context name and straxen version, we can get the hash for each dataype. For example, for the xenonnt_online context and straxen version 0.11.0:

>>> db.get_context('xenonnt_online', '0.11.0')

{'_id': '5f89f588d33cced1fd104ea5',
 'date_added': '2020-10-16T19:33:28.913000',
 'hashes': {'aqmon_hits': '4gwju6gdto',
            'corrected_areas': 'wbgyvcbq7i',
            'distinct_channels': 'a6orjqoffa',
            'energy_estimates': 'plxyjbnnui',
            'event_basics': 'pjql7a36yb',
            'event_info': 's4mprxr7qq',
            'event_info_double': 'hodh2726fi',
            'event_positions': 'tzzwhaf4gy',
            'events': 'aycpwrvco2',
            'hitlets_nv': 'pd2fuwzbjt',
            'led_calibration': 'lsdigsccxn',
            'lone_hits': 'nagx3zzuiv',
            'lone_raw_record_statistics_nv': 'vx2cbuxcxo',
            'lone_raw_records_nv': 'vx2cbuxcxo',
            'merged_s2s': '274qjtnjto',
            'merged_s2s_he': 'n5tib6rljj',
            'peak_basics': '2gq3vm6b2t',
            'peak_basics_he': 'ucjr5lvvbg',
            'peak_positions': 'c7vzetbnqw',
            'peak_proximity': '7nvjls2mab',
            'peaklet_classification': 'gpnp6dzxc4',
            'peaklet_classification_he': 'tmlx5dkfcf',
            'peaklets': 'nagx3zzuiv',
            'peaklets_he': 'mu26cr25vf',
            'peaks': 'dmavalropc',
            'peaks_he': 'hzamxeoer6',
            'pulse_counts': 'jxkqp76kam',
            'pulse_counts_he': '5mepav2zzf',
            'raw_records': 'rfzvpzj4mf',
            'raw_records_aqmon': 'rfzvpzj4mf',
            'raw_records_aqmon_nv': 'rfzvpzj4mf',
            'raw_records_coin_nv': 'vx2cbuxcxo',
            'raw_records_he': 'rfzvpzj4mf',
            'raw_records_mv': 'rfzvpzj4mf',
            'raw_records_nv': 'rfzvpzj4mf',
            'records': 'jxkqp76kam',
            'records_he': '5mepav2zzf',
            'records_nv': 'btmwmjvgdp',
            'veto_intervals': '7q35r23nba',
            'veto_regions': 'jxkqp76kam'},
 'name': 'xenonnt_online',
 'strax_version': '0.12.2',
 'straxen_version': '0.11.0'}

If you know the specific datatype whose hash you need, use instead get_hash:

>>> db.get_hash('xenonnt_online', 'peaklets',  '0.11.0')
'nagx3zzuiv'

If you are deemed worthy to have write permissions to the runDB (you have a corresponding user/password with write access in your config file), you can also add documents to the context collection with

>>> db.update_context_collection(document_data)

where document_data is a dictionary that contains the context name, straxen version, hash information, and more as shown in the example above.

Boilerplate pymongo setup

The runDB API is the recommended option for most database queries, but sometimes a specific query isn't supported or you might want to do complex aggregations, etc. For that reason, utilix also includes a wrapper around pymongo to setup the MongoClient. To use this, you need to specify in your config file

[RunDB]
pymongo_url = [ask someone]
pymongo_database = [ask someone]
pymongo_user = [ask someone]
pymongo_password =  [ask someone]

Note that this is needed in addition to the runDB API fields. Given the correct user/password, you can setup the XENONnT runDB collection for queries as follows:

>>> from utilix.rundb import pymongo_collection
>>> collection = pymongo_collection()

Then you can query the runDB using normal pymongo commands. For example:

>>> collection.find_one({'number': 9000}, {'number': 1})
{'_id': ObjectId('5f2d999448350bff030d2d3b'), 'number': 9000}

You can also access different collections by passing an argument to pymongo_collection. The analogous query to the contexts collection shown above would be:

>>> collection = pymongo_collection('contexts')
>>> collection.find_one({'name': 'xenonnt_online', 'straxen_version': '0.11.0'})

If you need to use different databases or do not want to use the information listed in your utilix configuration file, you can also pass keyword arguments to overwrite the information in the config file. This is useful if you need to e.g. use both XENONnT and XENON1T collections. To use the XENON1T collection, for example:

>>> xe1t_coll, xe1t_db, xe1t_user, xe1t_pw, xe1t_url = [ask someone]
>>> xe1t_collection = pymongo_collection(xe1t_coll, database=xe1t_coll, user=xe1t_user, password=xe1t_pw, url=xe1t_url)

Data processing requests

You may find yourself missing some data which requires a large amount of resources to process. In these cases, you can submit a processing request to the computing team.

If you have utilix installed, your context will have the request_processing method.

Example:

    st = straxen.test_utils.nt_test_context()

    requests = st.request_processing('012882', # can be a tuple of run_ids
                                     'event_basics', # must be a single data type
                                     priority=1, # higher values mean higher priority, defaults to -1
                                     comments='this is for the new elife analysis', # add some comments to help people reviewing the requests.
                                     submit=True, #If you use `submit=False` the requests objects will be created but no submitted.
                                     )

The script will open a web-browser if possible, and also print out the URL you need to visit to authorize the script to operate on your behalf. After you login using github and authorize the request, the script will get a token and save it in memory for the next requests.

Batch queue submission

Class JobSubmission (Since v0.8.0)

The JobSubmission class is designed to simplify the process of creating and submitting jobs to a computational cluster. This class encapsulates all the necessary details required to submit a job, such as the command to execute, job logging, partition selection, and resource allocation. It's built to be flexible and easy to use, catering to a variety of job submission needs.

  • jobstring: The command that will be executed by the job. It is mandatory.
  • log: The filepath where the job's log file will be stored. Default is "job.log".
  • partition: Specifies the partition to submit the job to. Supported partitions are dali, lgrandi, xenon1t, broadwl, kicp, caslake, build. Default is xenon1t.
  • qos: Quality of Service to submit the job to. Default is xenon1t.
  • account: The account under which the job is submitted. Default is pi-lgrandi.
  • jobname: The name of the job. Default is somejob.
  • sbatch_file: Deprecated.
  • dry_run: If set to True, the job submission will only be simulated. Default is False.
  • mem_per_cpu: Memory (in MB) requested per CPU. Default is 1000MB.
  • container: The name of the container to activate for this job. Default is xenonnt-development.simg.
  • bind: A list of paths to add to the container. It's immutable for the dali partition.
  • cpus_per_task: Number of CPUs requested for the job. Default is 1.
  • hours: Maximum duration of the job in hours. Optional.
  • node: Specific node to submit your job to. Optional.
  • exclude_nodes: List of nodes to be excluded from submission. Optional.
  • dependency: List of job IDs that must complete before this job begins. Optional.
  • verbose: If True, prints the sbatch command before submitting. Default is False.
from utilix.batchq import JobSubmission

job = JobSubmission(
    jobstring="python my_script.py --run_list 123456 234567",
    log="my_job.log",
    partition="xenon1t",
    qos="xenon1t",
    jobname="my_analysis_job",
    dry_run=False,
    mem_per_cpu=2000,
    container="xenonnt-2024.01.1.simg",
    cpus_per_task=4,
    hours=12,
    verbose=True
)

job.submit()

Method submit_job (Legacy)

If you have a script written with utilix<0.8.0, you probably used the submit_job method. This method is now deprecated but still available for backward compatibility. It's recommended to use the JobSubmission class instead.

The example below is equivalent to the previous one:

from utilix.batchq import submit_job

submit_job(
    jobstring="python my_script.py --run_list 123456 234567",
    log="my_job.log",
    partition="xenon1t",
    qos="xenon1t",
    jobname="my_analysis_job",
    dry_run=False,
    mem_per_cpu=2000,
    container="xenonnt-2024.01.1.simg",
    cpus_per_task=4,
    hours=12,
    verbose=True
)

TODO

We want to implement functionality for easy job submission to the Midway batch queue. Eventually we want to do the same for OSG.

It would be nice to port e.g. the admix database wrapper to utilix, which can then be used easily by all analysts.