UT1 TSE Calendar Sync and Filter

A tool to synchronise personalised Toulouse School of Economics (or UT1) timetables to your calendar app

Web page syncing to phone calendar app

1 September 2024

This web page is to help students at the Toulouse School of Economics (TSE) synchronise their class timetable to their phone (or Google calendar). This will work for all Université Toulouse Capitole (UT1) students (such as TSM).

You can normally view all classes for your course in a browser here. However sometimes classes are cancelled within an hour of them happening. Syncing to your device would be nice to help you notice this. Additionally, the web page calendar includes electives you're not taking. Most students can view a somewhat personalised timetable here. However that includes TPs, TDs etc which are not your ones. (Erasamus students don't even have that.) The purpose of this page is to generate a new URL to put into your calendar app, such that you can see all of your classes, and no classes for other people.

The way this works is that you type in the names of the courses you're taking below, and this website will generate a new URL to put into your calendar app, which will be the the same calendar as here, but with unwanted events hidden.

Another issue is that the university calendar website is quite unreliable. Often it just won't load for a few hours. With my server in between your phone and the university's unreliable server, I can fetch updates frequently (on behalf of everyone), and cache the results so it's reliable when your phone tries to sync.

Degree Selection

Which course are you enrolled in?

If you are taking courses from multiple universities, e.g. mostly TSE but one TSM course, then complete this whole page twice (once for each university) to sync multiple URLs into your one calendar app.

Subject Selection (Inclusion Expressions)

Now type in the name of your courses, one per line. Events in the public calendar which do not match at least one of these will be deleted when your device downloads the calendar.

The filtering will be case-insensitive. Only part of the course name needs to match for the event to be included. (e.g. "Econometrics" will match "Applied Econometrics") For courses like IO, write "Industrial Organisation" on one line, and "Industrial Organization" on the next. (Both spelling variations appear in the timetable.) Similarly, check carefully for spelling mistakes. (e.g. the "Datanomics" course actually appears in the timetable as "Danatomics". To be safe you should include both.)

There are some events in your calendar which aren't for a specific subject, but are instead general workshops for everyone. (e.g. The welcome talk on the first day of the university year.) To include these, add one of the lines which appear in the descriptions of those events. For example M2 EA EA.

If you leave this empty, all events will be included (unless they match the exclusions below).

TD/TP Selection (Exclusion Expressions)

Next you will filter out TDs/TPs that you aren't taking, for the courses which you are taking. TDs and TPs are the light blue coloured classes on the timetable web pages. To find out which TDs/TPs are yours, visit your personalised timetable. (If you don't see any, skip forwards a few weeks.) Now compare that to the public timetable web page. If your courses are small enough that there are no TPs/TDs which you aren't enrolled in, then you can skip this section. Otherwise, hover your mouse over each TD/TP. They typically have a few random-looking text identifiers, like "EMABA12TD3". (These tend to end with "TD" or "TP") For each of the other TD/TPs in each of your courses, find such an identifier. Check that the TDs/TPs which aren't yours don't have that identifier. Now paste those identifiers into the field below. Any TP or TD without any of these in the description will be removed from your calendar.

If you leave this blank, all TPs/TDs will be included (as long as they match the course names from further above.)

Results

Your URL is:

Installation Instructions

This URL is not a normal "CalDav" calendar sync link, it's actually a .ics file download (because that's what UT1 offers). So some apps (e.g. your browser) will try to download it as a file, and then from there you can import it into a calendar app. The problem there is that this is a once-off process, and future updates to your timetable by the university will not be synchronised. To ensure future updates will be synchronised, try the following steps, based on your chosen platform:

No authentication is required. (It's a public timetable.) You do not need to enter your UT1 password.

If your calendar tool lets you choose the synchronisation frequency, feel free to choose some very frequent polling frequency (e.g. 30 minutes). The requests go to my server, not the university. My server will just serve cached results, which it updates from the university every few hours. (If this service becomes popular I'll increase that to hourly. But even if 100 students use this, only 1 request gets sent to the university server per course. Their server is super slow and unreliable. But I've worked hard on caching, retries and exponential backoff to mitigate this for the end user, whilst minimising load to their server.)

Note that if you want each subject to appear as a different colour in your calendar app, you can use this web page to generate a unique URL for each subject, using each subject name as the keywords. Then once you have many calendars in your calendar app, set a different colour for each one.

Events will be deleted from the calendar once they are a day old. So you cannot see past classes. This is the behavior of UT1's server. I can't change that.

Don't put too much faith in this tool. It might be imperfect. e.g. for exams, check your schedule manually in a web page (or the emailed PDFs.)

How To Make Changes

If you have been synchronising successfully, but then you realise that you need to modify your keywords, you can paste your URL from last time into the following field. Then the text fields up the top of this page will be filled in based on your previous response. Then you can adjust, and copy the updated URL into your calendar app like before.

Privacy and Security

UT1's calendars are public anyway. That's what I use (not your personal calendar). You do not need to give your password to me to use this service. This service gives you a filtered version of the public calendars.

Technical Implementation Details

If you're curious how this works, here's an explanation. The code can be found on GitLab.

UT1's public calendar web page is here. Inside that URL there's a part like showOptions=false. If you change that to showOptions=true then new buttons appear down the bottom left, which you can use to generate the URL for the full, unfiltered calendar. You can pop this straight into your calendar app if you want. But the UT1 server is dreadfully unreliable. It fails (e.g. 500 errors) for at least 1 in 4 requests. Even when it works, it's very slow (up to 30 seconds to download the calendar, which may be longer than the client's timeout configuration.)

The URL generated by this web page does not start with https://ade-production.ut-capitole.fr. Instead it's like https://something.lambda-url.eu-west-3.on.aws. This domain points to my server running in Amazon's cloud (AWS). My server in the middle is a proxy which caches the calendar file, and performs filtering of the events within it on the fly.

Except it's not a traditional server, it's an AWS Lambda Function exposed with a function URL. This means that instead of paying for a server to run 24x7, I just give Amazon a zip of my python scripts, and when someone's device accesses the URL, they assign a server from their fleet to run my code, serve the response, and then that server is deleted, all within hundreds of milliseconds. I don't have to worry about patching, scaling etc.

There's a second lambda, which polls the UT1 server for the raw calendars (scheduled with a Step Function, to have very slow backoff and retries, since the UT1 server is so unreliable.) This saves the raw calendars as .ics text files in AWS S3, as the cache. (We can download the calendar file now. But maybe later when a user requests it, the UT1 server will be down. Then I'll just serve from the cache.) There's a lifecycle configuration rule to delete all files in S3 after a while. (This is how I erase stale entries from the cache.)

Every time someone requests a new calendar (e.g. the first M2 PPD student) the raw ade-production.ut-capitole.fr URL will be added to DynamoDB. Then next time we poll the server, a Step Function will loop over these DynamoDB rows to figure out which files to download. These rows have a TTL value, that gets updated every time a request is made. Thus if all users of a given calendar (e.g. all M2 PPD students) stop using my service, after a few weeks I'll stop polling the UT1 server for that calendar, to reduce their server load. (And my cloud bill, but that's negligible.)

Note that by caching the unfiltered files and filtering on the fly, we can do a single download to serve multiple users. e.g. if you have many calendar URLs, one for each subject, your device will make a request for each, which will all hit the same cache in S3. Thus the UT1 server only gets one request per user per course (e.g. all M1 students), not one per user or one per user per filter keyword set.

The project is deployed with AWS SAM, composed of a few "serverless" resources. I haven't used SAM before. It's surprisingly easy to use. I've found that AWS Lambda URLs are shockingly simple and cheap. (Unpopular opinion: Lambda URLs are no less secure than using API Gateway or a custom WAF. You can even add IAM authentication if you want.) I have aggressive maximum concurrency and timeouts set on the exposed lambda to mitigate DOS attack cost. The application code is written in Python. This web page is all lightweight hand-crafted vanilla HTML, CSS, JavaScript (No frameworks) hosted for free with GitHub pages. (This web page just does string manipulation to generate the URL. It doesn't actually interact with the Lambda functions and python code.)

Conclusion

If you have any feedback about this web page and service, please let me know. I'm Matthew Davis, the Australian M2 student who asks lots of questions in class. I used to do this kind of coding for a living. That's how I know how to do this. Yes I spent more time writing this than I saved by not manually checking my timetable in a browser each day. But I enjoy coding, so this time wasn't a 'cost'.

“I am rarely happier than when spending an entire day programming my computer to perform automatically a task that would otherwise take me a good ten seconds to do by hand.” - Douglas Adams