How can I move object in the Flow?

[rtaylor]

Basically, how can I tell DSS that certain objects in the flow should retain a given relative or absolute position? Example: I have a project with ~20 folders in one part of the Flow. Each folder is named "name_YYYYMMDD", where name is constant and the date is the date of the contents of the folder. Once I connect these via recipes to the main section of the Flow, DSS appears to reorganize them randomly, not in alphanumeric order. The down side to this is if I need to update the folder contents, or verify it, I cannot move through folders sequentially by date or any other logical method due to the automatic positioning applied.

Ideally, folders would be placed in alphanumeric order, or could be placed or "anchored" in some absolute or relative relationship to other objects in the flow.

[UserBird · Accepted solution]

Hi,

It is not possible to manually move the objects around in the flow. Perhaps you can switch to Anchoring mode: https://www.dataiku.com/learn/guide/tips/best-practices.html#flow_management

Cheers, 

Du 

[rtaylor]

Du,

Anchoring helps a bit from a sprawl perspective, but does not really address the positioning issue in complex Flows. Regarding the actual placement of items, is there any plan to look at ways to let users adjust the positioning in the Flow in future versions? I know this has been a bit of an annoyance for myself and some colleagues, as we cannot depend on the visual representation to remain consistent as we build and modify Flows in Projects.

Here are some examples where the dynamic, automatic nature of the visual Flow can become a bit of a issue:

Presenting a Flow to a group. If the flow was one way while you worked on it, and then it changes orientation, this can be troublesome during a screen share of if someone looks at a slide deck with an image of the flow then gets something different when they actually open it.

While building a new project. While adding steps to a flow, the flow can re-arrange in sometimes dramatic ways. This can cause some mental jarring, especially when the whole thing reorients, and especially when you have flows that move beyond a single screen comfortably.

Thanks!

[cperdigou]

I really encourage you to look into tags. You can use tagging to identify different parts of your flow. Even if it moves around the colors will really help.

Another good practice is to split your work into several projects. You can share datasets between projects. (usually the output of one being the input of another). https://doc.dataiku.com/dss/latest/security/exposed-objects.html#exposing-objects-between-projects

[rtaylor]

I agree flags have a place, but they do not resolve the issue that the unalterable nature of the Flow prevents users from applying any sort of organization via the orientation of items.

I agree that splitting projects may be viable in some cases, but if an entire Flow represents a single logical set of actions, splitting it into multiple projects just to organize it seems a bit backwards and likely to cause other problems, such as properly tracking a set of connected projects. For example my DSS instances (3 months and 9 months old, respectively) currently have 40 and 178 Projects, with more being added by the day/week.

Anecdotally: Almost every user I have introduced to DSS has been baffled and frustrated by inability to re-position objects in the flow, and this response usually happens when the flow decides it needs to rearrange a layout.

[florianbriand]

+1

There are some DSS's competitors who allow to add text or visual blocks in the "workspace".
This is useful to help on collaboration, identifying subparts of the flow, of with labels on the "wire" between 2 datasets, to explain in details what's going on.

From my point of view, this is one of the "biggest default" of DSS