Cache survey results #3201
Cache survey results #3201
Conversation
This comment has been minimized.
This comment has been minimized.
|
OK, I've fixed the problem by mimicking the caching we do for the catalog (caching at the template level instead of the python level) to avoid any user-specific or form-specific parts. I'm still thinking about how to cache the page for teachers or for when we look up a teacher, but really those have always loaded pretty fast, so I'm not really sure we need to (~1500-2000 ms of CPU Time on my dev server). |
willgearty
changed the title
This comment has been minimized.
This comment has been minimized.
|
LGTM alerts are fine, that's how that file works (the unused imports make the caching work). |
|
OK, I decided that we could do better. Now we cache each part of the survey results individually so we only refresh those parts when they have new relevant answers (the entire "individual questions" part is one part, and then each section is a separate part). So, for example, if a student fills out a survey for a particular section, only the cache for the part of the page that shows the results for that particular section gets refreshed. Since this means we have a little more rendering to do each time the page is loaded, the minimum time to open the page is now a little longer, but the average time to open the page should be a lot shorter, especially during times when students are occasionally filling out surveys. This also means that we can cache the parts for teachers (and when we look up specific teachers), and the really cool part is that if a teacher has recently opened their page, the parts of the survey review for that teacher's sections will already be cached if an admin opens the full review (and vice versa). Example times now:
Note: I believe the biggest component of loading the survey review page is creating and loading the histograms for numeric questions. The more histograms in a survey component, the longer that survey component will take to load if the cache is refreshed. |
This comment has been minimized.
This comment has been minimized.
* Initial docs for stable release 13 * Docs for #3116, #3117, and #3118 * Added docs about django upgrade * Docs for #3128 * Docs for #3129, #3133, #3134, and #3137 * Docs for #3156 and #3153 * Docs for #3174, #3163, and #3184 * Docs for #3139, #3140, and #3141 * Docs for #3143, #3150, #3154, #3160, #3162, and #3168 * Docs for #3171, #3185, #3186, and #3188 * Docs for #3131 and #3189 * Docs for #3149 and #3190 * Docs for #3193, #3194, #3195, #196, and #3197 * Clarification * Docs for #3192, #3201, #3209, and #2248 * Docs for #3204, #3212, #3214, #3205, 9fd073c, and #3226 * Docs for #3232, de5861c, #3231, #3233, #3234, #3237, #3238, and #3239 * Fix indent * Docs for #3227 and #3235 * Add missing word * spelling * Docs for e57581f, #3255, #3253, #3257, and #3249 * Docs for #3254, #3260, and #3262 * Docs for #3263, #3272, #3240, #3264, #3266, and #3270 * clarifications * Docs for #3283 and #3252 * Docs for #3288 and misc commits * Docs for #3292, #3311, #3286, #3289, and #3279 * Docs for a377f0d; move note * Docs for #3315, #3290, and #3322 * Docs for #3273 and #3317 * Final edits
This caches the survey review pages for admins (html and pdf versions!), making them load a lot faster once they've been run once. The rendering for each survey is cached independently, and we only ever update the caches when new responses are saved (just like how each class on the catalog is cached independently).
Examples of speed up (on my dev server, using "CPU Time"):
Fixes #3181.