💾 Archived View for thebird.nl › gn-gemtext-threads › topics › testing › top-10-tests.gmi captured on 2023-04-19 at 23:29:14. Gemini links have been rewritten to link to archived content
⬅️ Previous capture (2022-04-28)
-=-=-=-=-=-=-
Important things:
- Correctness
- Speed (fast enough)
- Reliability
Let's do 1 test at a time and run that by everyone. If we can one test a week we'll hit good coverage soon. Preferably, let's start with "Search".
1. Computations from Genenetwork2
For a test-case, consider using "UMUTAffyExon_0209_RMA" (full name "UMUTAffy Hippocampus Exon (Feb09) RMA") which appears under "Mouse -> BXD -> Hippocampus" in the drop-down. This is a very large data set. It might be computationally more efficient to start with a smaller data set: the 45,101 row-equivalent "Hippocampus Consortium M430v2 (Jun06)" data set, and then if all is good then test using the aforementioned exon-array monster [1,236,086 row-equivalents]; a 27X larger data set. Timing differences would be interesting to note.
2. "Local"/ "Global"/ "Wild card" Search
Test search with some horrid strings to see how resilient Search is to odd characters and those that MariaDB may want to treat as special strings: extra spaces, underscores, semicolons, Greek letters, tabs, etc.
3. Collections
Collections should involve user or code-generated synthetic traits, such as the residuals of a partial correlation, or an eigenvector trait. This is actually a key feature of network analysis-- part of the dimensional reduction game. Collections should also include genotypes. We need to check behaviour if the same trait vector values are entered twice under different (or the same) ID since this can torture the code for computing correlations such as "All-zeros vector?"
4. Data Tables With Huge Results Set
5. Correct Menu generation (Main Page)
6. Editing Data (Published Phenotypes)
7. Broken [External] Links
8. Working Jupyter Notebooks
This should be reliable enough. Use Case/ Perspective: Students are given assignments.
9. Authentication