Disclaimer: This is my first post about code ever. Any advice/tips are hugely appreciated and it's probably irrelevant and incoherent for most users, however, I was writing this down to better understand the problem myself so I might as well share it. If you don't like it or don't agree, let me know!
Time has multiple sources, solving strange time processing problems when using different ORM's.
Before I calculate and compare times, I should make sure that sources are aligned!
Otherwise the comparisons aren't right.
I always find it hard to reason about time.
Time is a very abstract concept to me, and it gets even more abstract when adding timezones and stuff.
So in this post I try to reason about time, because I had a problem.
I made a time comparison between a db-value that got sent to the client and processed there, and the same db-value, processed the same way, on the server. Somehow, the processed versions of that timestamp were not the same! How could it be?
As a very general start, I want to be clear about all possible sources from time.
I am currently aware of the following possible sources:
A - Local time of the user (set in the user phone settings)
B - Server time, as requested by node
C - Server time, as requested by a MySQL Query
D - Server time, as requested (on the client side, from the server) by GraphQL and then handled in React Native
What How A Date.now() New Date(Date.now()) B Date.now() New Date(Date.now()) C1 query('SELECT timeProperty FROM table;').then(obj => obj.timeProperty) 2 UPDATE table SET timeProperty = NOW(); D data.timeProperty
Some background of my problem: My algorithm that got some trouble:
The problem was just facing, is that I use time functions to validate a request to be a valid one. For a request to be valid, it has to contain this hash that's using a secret that is the same on the client and the server. I create this hash by encrypting a string containing a secret and a timestamp on the client. The timestamp on the client is gotten from the server just like option C1. Then, once the hash is sent to the server, the server creates a hash, based on the same database property value, using the same secret. The hashes are then compared to eachother. If they are the same, the request is valid because nobody knows the secret, except for the client (the app) and the server. If you don't know the secret, you can't make a valid request.
There are a few components in this formula.
- One is the hashing algorithm. I'm using sha256, which is v0.9, which is working the same on the client and the server.
- The second is the secret. It's a simple string, which is the same on the client and the server.
- The last one is the timestamp. It has the same source, but it is processed in a different way. It seems to be problematic because once they are compared they are not the same anymore.
hash = sha256(secret + date);
How do they get processed?
On the server, it's simple. Just query it from the database. However, in my case it's not that simple, because it can be done in two different ways: Using Sequelize, or using MySQL2 to perform the query. I am using two different ORM's on the server! And that was the problem! I used mysql2 in the resolver used to get the time on the client, while I used Sequelize to validate it on the server. And the results are different!
This is an example taken at 12:15 today in Berlin.
I performed the query
UPDATE users SET actionAt=NOW();
Then looked at what I would get when I queried actionAt.
user is requested using Sequelize, while user2 is requested using the mysql2 library.
user.actionAt: Sun Feb 04 2018 13:15:29 GMT+0100 (CET) user2.actionAt: Sun Feb 04 2018 12:15:29 GMT+0100 (CET)
Strangely enough, sequelize got it wrong!
Looking at GitHub, I found this is said by a maintainer of Sequelize:
"Your diagnosis of the problem is correct - the problem is exactly that sequelize converts the date to UTC, but when its returned, node-pg converts it into a date in the local timezone of the machine. I'm going to fix this in #4186, which allows for a more abstract way to hook into the parsing system than playing around with pg-types"
So it's probably converted to UTC time and then it's assumed to still be GMT+0100 while not having that. That's why the time goes up by one hour when using sequelize.
If I perform the same thing on the remote server (which sits in the London timezone), I get this output (when I'm in Berlin at GMT+0100):
user.actionAt: Sun Feb 04 2018 11:46:23 GMT+0000 (UTC) user2.actionAt: Sun Feb 04 2018 11:46:58 GMT+0000 (UTC)
Now there is no problem because the assumption of the date being in UTC format is correct because the server is in London! Here is some background information on what Sequelize does with its time: https://github.com/sequelize/sequelize/issues/854
My fix and opinion
To fix this when the assumption IS indeed incorrect, Sequelize advices me to set the timezone to the correct one when defining the sequelize connection. However, I don't want to do this! Because my timezone on the local server is different from my timezone on the remote server, and whenever I change my server to another server somewhere else, it will be the incorrect timezone and I'd have to change the script again! Using mysql2 directly, this is all not needed.
After using mysql2 in both cases, the problem was solved!
Lessons Learned: ORM's only make things more complicated! In the case of Sequelize, it generates flexibility but also adds another layer of complexity to your data! Avoid it if it doesn't give you a clear advantage, or you may have to add extra boilerplate to get your data right!