ultima-ib / ultima Goto Github PK

"Summary" to be availiable as a little button at the bottow right. Notice, what we can do: Summary table to show those statistics for each column (rather than just one selected)

Currently to call .ui() in rust one needs to build ultibi_server which requires

1. npm run build frontend
2. Set env variable STATIC_FILES_DIR
3. And only then build rust lib

This could (in theory) be performed from build.rs like this:

1. Git checkout ultima/frontend
2. npm run build
3. call to resource_dir with dist forlder as an arg

Currently, user builds request and Posts it by hitting "Run". This is known as Deffered Update (ie only update after hitting Run).

We could improve performance significantly, by sending "incremental" requests. Say, wants to add a single measure to his request (and single measure only). What we could do, is in fact send a request just for that additional measure as soon as user has selected it, and then outer join responce to the existing result !on GroupBy fields!. Polars nodejs has all the built in functionality. It's just about applying it in our context.

Clearly, this only works if user has selected at least 1 Group By field. It also only works if user hasn't changed Group By, or Filters, or Overrides. If they did, then we do not send incremental request.

Additional backgroupnd about outer join.

It would be good to do anything to mask this void space. Its purely visual. Functionality is correct

• "Use custom template" button could contain the existing request (so user can copy request from there, or copy paste the request from somewhere else)
• Remove Copy Requst Button.

@hamza1311 , would it be possible to:
i) Highlight Column of focus same way as you highlight row
ii) Make the cell value of focus bold
ii) Use accounting format for numbers, ie when number is 997895.11 we should show 997,895.11 This is much more user friendly.

• Highlighting Total lines of the result table
• Highlighting row/column on hoover
• Borders across measures box And fields box (at the moment user might get confused on what is what on the most left hand side)

It's possible clients will have too much data to hold it in memory all the time. For such usecases we can try to provide an option for lazy scans, leveraging on with_streaming.

Pros:

1. No in memory frame

Cons:

1. Performance will be impacted significantly because reading, assigning weights, has to happen per request.

Polars bug. TODO: raise to polars

It's super simple. Just login to aws, go to IAM, and you will see in red the link to do it.

The idea is that a user can convert each table into a chart.

Next to Summarize add Charts
At the top left corner of ech table a small seethrough (graph) icon button.

When user clicks, eather a dropdown or more buttons appear. These follow the hierarchy of chart.js samples. In our case user will need to click "Other Charts", this will open more buttons. Then "Pie".

Attached document Pie tab shows the output for Result from Data tab.
Charts.xlsx

Note: the nature of pie chart is as such for each Measure we will get a Chart. Show them next to each other. Keep in mind it can be 1,2,3,4 etc. Let user close the ones he/she doesn't need.

It would be really good if you could let user perform actions such as Add Labels (in my document Labels are added).

https://docs.rs/actix-web/latest/actix_web/struct.App.html#method.configure

eg. I cannot select ("MeasureA", "sum") and ("MeasureA", "count"), but this should be possible unless Aggregation is restricted to only one possible value.

So once "MeasureA" was selected as "sum", it should reappear/remain in the list, and I should be able to find it again and choose "count" or "mean" or whatever

For Measure, in theory we could have MeasureA-sum and MeasureA-count. However, for now we can keep it simplier. Just make sure that the measure is unique in the request.

We can and should be able to override columns of types: DataType::List(DataType::Float64), DataType::Float64, DataType::Utf8 (and bool to be added) as per overrides.rs

There are two ways to achieve this:

1. Change returned fields in Dataset Info GET API endpoint to return column name and column type. Then FrontEnd to pick Utf8 for fields, and types mentioned above for overrides.

2. add another key (name it "overrides) into the return of the Dataset Info GET API endpoint, called overrides which would return column names of columns with above datatypes. This key (ie "overrides") then would be used for overrides (instead of fields).

@hamza1311 which one do you prefer?

A newly created API endpoint http://localhost:8000/FRTB/templates returns a vec of Requests. (Notice they have a new optional field called name). It would be nice for user to search and select from a dropdown for a template. The idea behind this is if an organisation has a template which many users use many times, it's would be much easier for them to select a template rather than type it in. (Moreover, organisation mught prefer users to use a particular template for particular purposes).

Equity bucket 11
DRC weights for CRR2
RRAO
CoveredBond
DayCountConvention for DRC yearfrac

User might want to compare two outputs from two different requests - side by side. (for exmaple, with filter vs without. Or with override vs without). For this, it would be beneficial to allow to "split screen", so that two result tables could be seen at the same time.

The required secrets and documentation is mentioned here: https://github.com/docker/login-action#aws-elastic-container-registry-ecr

They're used at:

https://github.com/AnatolyBuga/ultima/blob/acb4ecacbefe6219055bc55cda79cfcffd95305d/.github/workflows/main.yml#L91-L96

Way to return an error:

pub fn error_expr() -> Expr {
    apply_multiple(|_|{
        Err(PolarsError::ComputeError("Could not deserialize".into()))
    }, 
        [col("*")], 
        GetOutput::from_type(DataType::Float64), 
        false)
}

Most likely due to added /dist/. folder (which needs to be added for docker to build).

Library to be called from Python

best for smaller clients

Difficult while polars doesn't allow that. Have to do it via python.

Must have for on the fly analysis

It would be really good, if possible, to place AggType next to the Measure name and then allow multiple measures in a row (similar to GroupBy)

P.S. This is just thinking how we could save some space

• filters
• overrides

Consider using join instead of when.then

Two digits after the . would be great. 1.9999 Ideally would be 2.00, but if we make it 1.99 that would work too (but 2.00 preferred).

Request to be broken down into incremental requests in the backend and then for each single measure in request
to check if such request has been posted. If so, look it up from the dashmap, otherwise send and save incrementally)

User to download result as a CSV file

Polars 25.1 introduced scalar param to apply_multiple function. This could be used to instead of creating a series of particular height.

i) Imagine, User A and User are connected to the same backend, ie they are within the same orginisation. If Anatoly right clicks on the table and sees an option to view request which opens a super simple text box with request body and copies and pastes it to clipboard. Then he/she email/message this to User B (this step is independent of us).

ii) Then User B pastes this request into the same text box on his side, and then presses Run/(or presses OK in the same box), and this sends the request to the server , and User B gets the same table on his side.

Hey Hamza,

After going through all the calc params, #13 , I see that some of them have to be really long. There is not much we can do here. It's jsut how the regulator (ECB) prescribes them. As such, I think it would be really good to use dropdown box with search (and removal) for this. What do you think?