Research DeepSee Rest API - instead MDX2JSON

[Lena-Ev]

• [ ] Methods list with possible realization MDX2JSON and DeepSeeRest API

DeepSeeRest API <- MDX2JSON <-DeepSee Web DeepSeeRest API <- DeepSee Web

Try to use both technologies at the same time

[Lena-Ev]

list of all the methods and classes

[Lena-Ev]

@jakcpto

[Lena-Ev]

@jakcpto check slow-loading dashboard ask Eduard

[Lena-Ev]

List of all mdx2json methods https://docs.google.com/spreadsheets/d/1Vp3LJ-L-zo87J6lViVy7AHutkWtgG0YXoGIZL01sj68/edit?usp=sharing

[Frisle]

Spreadsheet were updated with time spend on run for each method. Used widget:

[Lena-Ev]

new test case - result set mdx-> result set DeepSee

[Lena-Ev]

https://analytics.community.intersystems.com/reports/index.html#/DC/IRIS_BI/En

[Frisle]

I consulted with Eduard about the topic. He suggested a profiling method to gather precise statistic from result set and after that he can help with picking the right method to replace the old one

[Frisle]

Here is a solution to optimize slow loading widgets. This method tries to figure out which cell should not be displayed. The commented lines are the ones that have been optimized.

The method was tested on a widget with 12000 rows. Before optimization load time was 30+ seconds, after ~7 seconds.

Method IsCellNull(Cell)
{
    #; Return:((Cell.type = "axis") || (Cell.type = "set")) 1 // for top-level cells

    return:(Cell.caption=0) 1 // typical caption for top-level cell
    return:(Cell.caption=1) 1 // typical caption for top-level cell
    return:(Cell.caption="") 1 // typical caption for top-level cell
   
    set children = Cell.children

    return:(($isobject(children)) && ($$$ListSize(children)>0)) 0 // cell has children
    
    return:(Cell.vis '= 0) 0 // in case of type='lit'. cell 'lit' is always vis=0 so we can skip it

    #; If (Cell.type'="lit") {
	#; 	Set key = $ORDER($$$DeepSeeResultsGLVN(..%CubeKey, ..%QueryKey, "leaf", AxisNumber, ""))
	#; 	While (key'="") {
	#; 		Return:(Node=$$$DeepSeeResultsGLVN(..%CubeKey, ..%QueryKey, "leaf", AxisNumber, key)) 0 //for leafs
	#; 		Set key = $ORDER($$$DeepSeeResultsGLVN(..%CubeKey, ..%QueryKey, "leaf", AxisNumber, key))
	#; 	}
	#; } Else {
	#; 	/// lit cell: SELECT %LABEL("Const","Title") ON 0 FROM [HoleFoodsBudget]
	#; 	/// but also a lot of top-pevel cells
	#; 	Return 'Cell.vis
	#; }

    return 1
}

[Frisle]

So progress has been made without implementing DSZ methods, these are purely tweaks in the result set. As @evshvarov pointed out the advantage of DSZ is now around 20 seconds in between API in case of very big pivots (2.35m DSZ vs 2.50m MDX2JSON). Should we continue attempt to replace MDX2JSON with DSZ or we should optimize Result Set futher?

[Lena-Ev]

DC report: Views dynamic by members post

[Lena-Ev]

add to Excel table what can be done with REST API and what can be done only with MDX2..? If it is possible then what are the obstacles? | REST API | MDX2... -yes,

Python?

[Lena-Ev]

duplication of functions of API additional functions to API gap of deepsee rest UI for deep see web

---

Estimation of work