There is a newer version of the record available.

Published April 1, 2017 | Version v3.0.0
Software Open

Dallinger/Dallinger: 3.0.0

Creators

  • 1. UC Berkeley
  • 2. Glick Software
  • 3. @pyupio
  • 4. The Code Distillery
  • 5. University of California, Berkeley
  • 6. Cris Ewing, Developer

Description

Welcome to Dallinger 3. This release comes with several new features, some of which are breaking changes that will require you to edit your .dallingerconfig file and experiment code. This changelog will be updated to reflect any new breaking changes that we discover.

Several configuration parameters have been renamed or removed. In particular, to migrate, you MUST:

  • Rename amt_keywords => keywords
  • Delete psiturk_keywords
  • Delete launch_in_sandbox_mode
  • Delete section [Shell Parameters]
  • Delete anonymize_data
  • Delete table_name
  • Delete psiturk_access_key_id from .dallingerconfig
  • Delete psiturk_secret_access_id from .dallingerconfig

Additionally, note that section headings are now optional, meaning that all configuration parameters must have a unique name. We recommend that you:

  • Rename [Experiment Configuration] => [Experiment]
  • Rename [HIT Configuration] => [MTurk]
  • Rename [Database Parameters] => [Database]
  • Rename [Server Parameters] => [Server]

The command dalinger verify should catch configuration-related issues.

  • BREAKING. When testing experiments locally using dallinger debug, recruitment is now automatic and does not require you to run debug in the psiTurk shell. The workflow for debugging an experiment used to be:

  • Run dallinger debug

  • Run debug in the psiTurk shell
  • Participate in the experiment
  • Repeat steps 2 & 3 as desired

The new workflow is:

  1. Run dallinger debug. This will directly open a new browser window for each participant that is recruited.

  2. Participate in the experiment.

  3. BREAKING. There are two breaking changes with regard to recruitment First, the recruiter's recruitment method has been renamed from recruit_participants to recruit. Second, the default recruitment method no longer recruits one new participant; instead, it does nothing. Thus to retain the 2.x behavior in 3.x experiments that do not override the default, you should include the original default recruit method in your experiment.py file:

def recruit(self): """Recruit one participant at a time until all networks are full.""" if self.networks(full=False): self.recruiter().recruit(n=1) else: self.recruiter().close_recruitment()

FEATURE. Addition of a high-level Python API for automating experiments and a data module for handling Dallinger datasets, making it possible run experiments in this way:

import dallinger experiment = dallinger.experiments.Bartlett1932() data = experiment.run({ mode="live", base_payment=1.00, })

FEATURE. There is a new data module, dallinger.data, which provides a few new pieces of functionality. First, you can load datasets that have been exported:

data = dallinger.load(UUID_OF_EXPERIMENT)

The returned object makes the dataset accessible in a variety of formats, including a pandas DataFrame and CSV file.

FEATURE. On export, data is automatically backed up to Amazon S3.

FEATURE. Integration with Open Science Framework. When an OSF access token is added, each experiment launched in sandbox or live mode will create a new project on the Open Science Framework and back up your experiment code in that project. We will be developing deeper integrations in the future.

Files

Dallinger/Dallinger-v3.0.0.zip

Files (2.9 MB)

Name Size Download all
md5:616a216ac0c3492d50d6617d8a0ea63b
2.9 MB Preview Download

Additional details

Related works

Is supplement to
https://github.com/Dallinger/Dallinger/tree/v3.0.0 (URL)