Event bots are generated from their base type, then their role is updated before being sent to the client e.g. peacefullZryachiyEvent becomes bossZryachiy for generation, then back to peacefullZryachiyEvent
- Implement formula based on 30 weapon repairs done on live
- Return the repair amount as `repairAmount` from `repairItemByKit`
- Add an additional `repairPoints` to `RepairDetails` to return the repair points used
- Update `repairAmount` references to `repairPoints` to keep old behavior
- Add new parameter to rewardSkillPoints that applies live-like level scaling
- Only give weapon maintenance XP when using a repair kit
This implementation comes with a "Crit Fail" and "Crit Success" mechanic to account for live sometimes randomly being -4 or +4 off from my estimated values. By default the chance of each is 10%, and they can overlap and cancel each other out
Spreadsheet of live repair data:
https://docs.google.com/spreadsheets/d/1-tR4WYelhZfKZ3ZDbxr3nd73Y60E1wQRjDWONpMVSew/edit?usp=sharing
Useful columns:
C: The amount of dura attempted to be repaired, this is used in the estimated skill calculated
G: Hand entered value of how much skill gain I actually saw on live (Multiplied by 10 for readability. So "3.2" would be "32")
J: The estimated skill gain, based on the calculation included in this merge request
K: How far off the estimated skill gain was (Negative implies we guessed high and the actual result was lower)
One thing of note:
I've modified all the existing references to `repairAmount` to be the new `repairPoints` when a repair kit is used. This is to keep the existing behaviour outside of my direct changes as much as possible.
However, this seems to be incorrect in some cases (For example, buff chance is repairPoints/maxDura, but repairPoints will go down the higher your int. I'm assuming this is meant to be repairedDura/maxDura). May want to update these references to use `repairAmount` once they've been confirmed to expect the repair amount instead of repair points used.
Co-authored-by: DrakiaXYZ <565558+TheDgtl@users.noreply.github.com>
Reviewed-on: https://dev.sp-tarkov.com/SPT-AKI/Server/pulls/164
Co-authored-by: DrakiaXYZ <drakiaxyz@noreply.dev.sp-tarkov.com>
Co-committed-by: DrakiaXYZ <drakiaxyz@noreply.dev.sp-tarkov.com>
- Created profileInsurance test fixture.
- Created ProfileInsurance test factory class.
- Improved many of the InsuranceController method tests by utilizing the factory and fixture. Still some left to do here.
- Adds the `date-fns` package as a development dependancy for easy date manipulation.
- Cleaned up some comments.
Bug Fixes 🐞
- Resolved an issue where `ItemHelper.getAttachmentMainParent()` was not returning null when it should have.
- Resolved an issue where `InsuranceController.populateParentAttachmentsMap()` would continue to process when the parent could not be found.
- Adds tests for the majority of the methods within InsuranceController.
- Simplifies the `InsuranceController.sendMail()` method so that the toDelete parameter is no longer required.
- Modifies `InsuranceController.findItemsToDelete()` to only process regular and attachment items when they exist.
- Modifies `InsuranceController.rollForDelete()` to return null in the event that the provided Trader can not be found.
Removed `parcel` and `babel` config files and dependencies, as the project isn't using these tools anymore.
Swapped `rome` for `biome`, `rome` is archived and won't be receiving updates anymore, `biome` is a community fork.
Co-authored-by: Pedro Silva <sparta14gaming@gmail.com>
Reviewed-on: https://dev.sp-tarkov.com/SPT-AKI/Server/pulls/163
Co-authored-by: TheSparta <thesparta@noreply.dev.sp-tarkov.com>
Co-committed-by: TheSparta <thesparta@noreply.dev.sp-tarkov.com>
Adds a single test for the basic functionality of the `PaymentService.payMoney` method. It's huge. I hate it. It works. The entire class needs to be refactored to be more testable. Each of these methods should be split up to only have (ideally) a single responsibility--It has about 6 going on right now.
It was just too damn slow... 25+ seconds for *two*. Profile information will have to be partial-mocked using the sections and properties that the tested method will use, and casted as the parameter type (`IPmcData`, for example).
Remove unneeded HttpBufferHandler
-----------
The old code processed each chunk of data as an entire request, which is not correct. It was observed split data after ~14600 bytes (on a 1 gig lan connection). I think it was worse on remote connections.
This was the cause of the "unknown compression method" and invalid json parse errors when saving the profile.
Co-authored-by: Decoy <redacted@example.com>
Reviewed-on: https://dev.sp-tarkov.com/SPT-AKI/Server/pulls/162
Reviewed-by: Terkoiz <terkoiz@noreply.dev.sp-tarkov.com>
Co-authored-by: ree <ree@noreply.dev.sp-tarkov.com>
Co-committed-by: ree <ree@noreply.dev.sp-tarkov.com>
- Made it so Watermark doesn't initialize on the constructor, this makes it possible for tests to use `@spt-aki/di/Container`;
- Removed unnecessary calls to `ErrorHandler.handleCriticalError`, all you really need to do is `throw new Error()` and it'll get caught automatically.
Basically the same, except it plays nicer with Typescript and ESM.
I have it mostly working, except for a type error:
`TypeError: Int32Array is not a constructor`
But I'm too damn tired it debug it at the moment.
