This report will be a little bit longer than normal cause it will comprehend the last week changes too. So let’s start.

  1. Status: We have been accompanying the status and behavior of the update_versions alterations on Cirlce to check if any further modifications or bug removals would be necessary, with some minor configurations to the deploy phase of the versions files, we could see the new approach working very well since it was merged. As an extra task from my mentors, I added a profiler for the auto_tick.py script to verify the usage of the bot process and check for eventual optimizations we can do.

  2. Abstract: For the new_update_versions update, we decided to switch the actual update_version function for the new one, as also update the relative test script which can be seen on this PR. With a big help of CJ and Matt we removed the configuration for the original CI update process, and replaced the update_versions subprocess with the new one. Which was actually removing the update upstream versions from run bot:

      - run:
   name: update upstream versions	
   command: |	
       cd cf-graph	
       source activate run_env	
       conda-forge-tick --run 2

    and update the populate versions process with the following modifications, which was just removing the previous main function for new_update_versions of cli.xsh file and replace for the main of the update_upstream_versions: ```python

    • run: name: populate versions command: | cd cf-graph source activate run_env conda-forge-tick –run 2 During the test week, we saw some missing updated files from the feedstock, so CJ commented with a possible solution, it was to reformat the way we were reading the feedstock `bad` information (follow the conversation [here](https://github.com/regro/cf-scripts/pull/1073)), this way we replaced this python with node_attrs[“payload”] as attrs: if attrs.get(“bad”) or attrs.get(“archived”):
  with 
```python
with node_attrs["payload"] as attrs:
    if (attrs.get("bad") and "Upstream" not in attrs["bad"]) or attrs.get(
            "archived",
    ):

Then we get it running smoothly for the next week, now we only need to look for some code optimizations (if needed) and look out an process workflow explanation. Which comes forth with the idea of an profiler for the bot process, I’ve opted for the cProfiles as it attends to our necessities and have a better usage in my opinion, so the basic alteration made can be seen with this explanatory code:

def function(*args) -> any:
    # start profiler
    profile_profiler = cProfile.Profile()
    profile_profiler.enable()
    # Do something....
    # stop profiler
    profile_profiler.disable()

    # human readable
    s_stream = io.StringIO()

    profile_stats = pstats.Stats(profile_profiler, stream=s_stream).sort_stats(
        "tottime",
    )
    profile_stats.print_stats()

    # get current time
    now = datetime.now()
    current_time = now.strftime("%d-%m-%Y") + "_" + now.strftime("%H_%M_%S")

    # output to data
    with open(f"profiler/{current_time}.txt", "w+") as f:
        f.write(s_stream.getvalue())

which bumps the profiler information to a folder called profiler and contains the information needed as the respective run time (for inspection). These Profiler LOG’s allows us to receive some feedback for the time usage of the process and think of better ways to optimize then.

  1. Next: For the last meeting with my mentors and the conda-forge bot team, we decided on some topics for better time and efficiency usage for the bot process, profiling results, as also discussed some future features for the bot structure. My tasks for next meeting will be:
    • run the profiler on the pr json update steps too;
    • make a PR to switch the bot to use rapidjson.

I also received the Google Summer of Code email for the last phase of the project and information regarding the final evaluation with the project resume and code work product. For this purpose I will also be working on the final work presentation during this last weeks.