While developing experiment-specific code, a developer may work on a feature (such as an algorithm, utility, improvement) that can be shared with the community of experiments that use LArSoft. In most cases, this can be done easily since the modified code does not affect the architecture and doesn’t break anything. But when the feature affects the architecture, a few more steps are needed to ensure that such contributed code can be properly shared across experiments and integrates well into the existing body of code. This article outlines the process to introduce and integrate into the core LArSoft code repositories both non-architectural features and architectural-affecting features.
The intent of the process is to achieve a smooth integration of new code, new ideas or improvements in a coordinated way, while at the same time minimizing any additional work required beyond the normal development cycle. Many changes will not require prior discussion. In cases with broad architectural implications getting feedback and guidance from the LArSoft Collaboration or the core LArSoft team early in the development cycle may be required.
Process for contributing a non-architectural, non-breaking change to LArSoft:
- Become familiar with the design principles and coding guidelines of LArSoft.
- Develop the code including comments, tests and documentation.
- Offer it by talking about it to LArSoft team members and maybe give a presentation at the LArSoft Coordination Meeting.
The rest of this article addresses a breaking, architectural change to LArSoft. Remember, all other cases will involve a more simplified process.
Process for contributing an architectural, breaking change to LArSoft:
- Someone working on an experiment has an idea, an improvement, or a new feature that affects the LArSoft architecture that can be shared in the core LArSoft repositories.
- Developer contacts the LArSoft Technical Lead or other members of the core LArSoft team to discuss the idea. Discussion can include an email thread, formal meeting, chat in the hallway, phone call, or any communication that works.
- May find that the feature, or a suitable alternative, is already in LArSoft, thus saving the need to develop it again.
- At this point, a decision will be made as to whether further discussion or review is needed, and where that discussion should take place. If other experts are likely to be required, a plan for including them in the process will be developed. The division of labor for the architectural changes will be discussed as well.
- Developer learns/reviews the design principles and coding guidelines of LArSoft.
- Developer prepares a straw proposal or prototype for the change.
- For major changes as determined in step (2), the proposal should be presented at the biweekly LArSoft Coordination Meeting. Depending on the feedback, more than one presentation may be useful.
- The developer writes the code, including comments, tests, and examples as needed, and keeps the LArSoft team informed on the status of work.
- Any redmine pages or other technical documentation should be written during this time as well.
- For new algorithms and services, an entry in the list of algorithms should be made.
- When development is completed, request that it be merged onto the develop branch since this is a breaking change.
There is a cost in making things workable for other experiments, but the benefit is that other experiments develop software that is usable by all experiments. The more this happens, the more all experiments benefit.
When designing LArSoft code, it’s important to understand the core LArSoft suite and all the components used by it. It’s also important to follow the design principles and coding guidelines so that what is developed can be maintained going forward. A good place to start is at Important Concepts in LArSoft and by reading the material and watching the video about Configuration. It is important to follow the guidelines to have configuration-aware code that is maintainable with easy-to-understand configurations. The less time that is spent on debugging, the more time that can be spent on physics.
Once LArSoft contributors are aware of how LArSoft works, including the principles and guidelines for developing software, and have discussed the new feature with the LArSoft team, they will usually be asked to make a presentation at the LArSoft Coordination meeting. The idea here is to share the plan and the approach to solicit more ideas. Treat the initial presentation as a straw proposal–something that is designed to solicit feedback, not the final implementation that must be defended. At the same time, if a suggestion would double or triple the work required to implement it, and there isn’t a strong need for that suggestion to be implemented, it can be noted and set aside. The contributor is in charge of what he or she implements. The goal is to share software to reduce the development effort overall. It also encourages communication across experiments. The more collaborations can benefit from each other’s work, the better off we all are.
Details on developing software for LArSoft can be found at Developing with LArSoft wiki page. By contributing to the common LArSoft repositories, improvements to the code can then be used by everyone.
— Katherine Lato & Erica Snider