CODAP Missing Projectile Data

[KatieWoe]

Device
Dell
OS
Win 11
Browser
Chrome
Problem Description
For phetsims/qa#1096
When a large number of projectiles have been launched, it is possible that some of them won't be picked up by CODAP. This does not happen every time, and so may be related to performance/speed/memory.

Steps to Reproduce

1. In Codap, go to the sim's first screen
2. Set sim speed to fast, and firing to continuous
3. Fire all the projectiles (500)
4. Once the shots are done, erase them and repeat
5. You should end this with 1000 data points, but some may not appear

Visuals

[KatieWoe]

Can be very bad in phetsims/qa#1099 when sample sizes are large.

[samreid]

Brainstorming a plan:

• See where the projectiles are getting lost. Does the wrapper know about them? If the wrapper knows about them but CODAP doesn't then consider throttling or batching messages from the wrapper to CODAP.

[matthew-blackman]

This has been fixed in the commit. Marking for review during RC 1.

[catherinecarter]

When I was testing the wrapper from another issue, I launched several projectiles in PSD+CODAP in this way:

• Same launcher (1)
• Launched with all four sample sizes
• Continuous and fast mode

I noticed not all the data came through. You can see in the screenshot that are a zero samples of size 40 in the CODAP table (look at the 2nd hierarchical variable), even though you can see on the field sign that there were 31 samples launched. The histogram also shows this data. I launched:

• 53 samples of size 2 = 106 projectiles
• 61 samples of size 5 = 305 projectiles
• 101 samples of size 15 = 1515 projectiles
• 31 samples of size 40 = 1240 projectiles

For a total of 3166 projectile. At the top of the table it says there are 996 cases, meaning there are 2170 cases missing.

I tried it again and was able to reproduce.

Then I tried it with each sample size, but waited until all the projectiles had been populated in the table prior to launching more with a different sample size.

• The table was correct for sample size 2 (43 samples – see the 3rd tier in the variable heirachy)
• Then I added sample size 5 (48 launches) without erasing any data, and the table was correct.
• When I switched to sample size 15, I launched 73 samples (for a total of 1421 projectiles launches between all sample sizes), but the table showed only 26 samples.
• I launched 30 samples with n = 40, the table tracked 24 samples

[catherinecarter]

I will note that for the samples CODAP did record, they had the correct corresponding number in them per sample size.

[matthew-blackman]

Good find @catherinecarter. This will be addressed in #349 when the queueing system designed for #338 is applied to the PSD CODAP wrapper.

[Nancy-Salpepi]

This looks OK to me with mac + chrome/safari in rc.1 but will keep open until other devices (Chromebook) and PSD are tested.

[KatieWoe]

I managed to get up to 7500 cases. After that nothing was recorded at all, so it seems like there is a max. Before that a lag started being noticeable but everything was being recorded. Win 11 Chrome