UI 2.0 - ja-guzzle/guzzle_docs GitHub Wiki
- Common Themes and Standards
- Other items
- Physical and Logical end point
- User Flows
- Samples and References
- Searching Tech for Connections
- UI Discussion
- Call on 10/Oct
- Call on 11/Oct
- Keep as much of metadata ready - column list and tables in schema etc - so that user can search for columns to work with; All the details are available - just type column and you got it - we can have all small (keep it seamless and not explicit)
- Gives shortcut to create things from whever.. - like in Tableau you can created workbook from Data source, or from the New workbook etc.; It just naturally lets you start thread from any where- example on logical end point and you can start creating jobs, duplicate a job..
- Fully bookmarkable URLs
- Back button support (wizard will support, but not when switching between functionalities /modules); => There is some handling, hvae to talk to @parth for it
- In a Job config edit session if user is siwtching technologies for end ponit or specific file types , or enabling /disable control file or watermark to retain the previous configs - Low priority : @umesh to confirm the requirement- for now its to prompt user that all the settings will be lost
- Provides simple undo - example i was in json and went to csv and again cam be back to json - all the context retained- how i don't know (it populates full object when we are editing). If you close and open. We don't support Control Z. Similarly we don't need UNDO for user actions like delete Low priority -> No Undo supported
- works on all the form-factor - though not super priority - Low priority
- No painful scrolling - specifically multi level scrolling
- Keyboard binding - Low priority => None
- Fluent https://www.microsoft.com/design/fluent/#/ - if we can follow as much : contrast /accessibitly, lighter tons good for eyes, any form factor, energy saving - Low priority
- It tells you whats happening - when you specify file path , it can you tell you what are the files matching - best if we can give count or initial list of 10 .. Feedback is must ; this includes also showing the effective mapping of source and target; You don't want user to guess whether they have done is going to work - they should know that its going to work - Low priority
- Default action for Single click on a particular object should lead to appropriate screen
- Do we need notion of data set like ADF - No
- Do we need a way to copy paste section of job configs - no
- Do we need a way to create jobs and job group together - like ADF pipeline ; No
- Import feature - do we need to change it as template - no
- Minimal data entry - example can we allow user to browse the SQL databases, file storage etc..
- Responsive UI - Look at Azure portal on how it handles diff resizing -without giving Horizontal - we are designing for fixed resolution
- Common Search at the top on all the metadata which is "default" or "branch" folder - the listing should show the object and section of matching config (like Google search)
- wrapping of long text (example SQL) or long list of jobs tagged to job group
- Tabbing between configs being edited simultaneously so that one switch among the configs - KIV
- Wizard based approach for editing configs (and not tabs)
- Ability to verify connections, job configs and get feedback the fastest
- Optimal fonts and spacing
- Language and Region selection - low priority
- Ability to choose color theme - low priority
- externalize all the boiler plates/labels
- Consistent way to give help
- Expression editor
- The text boxes can they exapnd when the content is huge , speically when we are focusing on on it or updating. Perhaps have a way to popup formual editor and there are three types:
- Glob + Groovy for file name pattern
- SQL + Groovy for Tranformation and vlidation (this gives you nice way to fit in SQL and groovy - bit complicated and foncusing sometime)
- Groovy alone for table name, partition columns etc
- For tables we can give list, for files we can give browse but it gets into the terroriotry of browing files in nested folder and hen later overriding with params or glob expression - once that is done it sort of gets delinked
- All the drop-down will be searchable - like Azure Portal =>
- Defaulting of things like compute or business dates- peole start ignoring and then they scramble on what happened - we have explict defaulting concept , we can bring to user session level instead of global
- Some where its Plus button : for context column, param , housekeeping etc; for tags its enter button - we have to agree on which => Follow the approach which we hvae for transformation everywhere when we are doing list. The current param and other ux - don't hvae edit option - so the only way is to delete and enter. So wiht this we solve few problem: . All that goes away
- Sampling - has to be come more stable in all the cases
- No right click
- When using back button on existing guzzle tabbed session;
- Launch a bokmark Url: http://spatest1.southeastasia.cloudapp.azure.com:8082/app/author/activity/job1
- When doing back or navigating out the unsaved changes will be discard - need it toprompt
- Duplicate tab of Chrome - should work like bookmark ; but most improtant two parallel guzle session in brwoser - can they sync
- With the same tech and sibling techs like jdbc , delta, hive- there are some liks sypanse and Azure sql connector when we design the ui thoes settings may be discarded ; remember or warn , mostly can't fit so thats fine
- Initiate task like getting file counts matching a pattern , or evaluating an expression on transformation and showing a sample value - the ux can get complex
- Save as or duplicate on left nav bar: we provide Save As when editing the job which prompts the job name of duplicate one (has to support for logical end point where it has button to duplicate physical end point which when selected prompts you the name of new physical end point).
- Delete on left nav vs Delete at the job level - we provide at job level
- if the help is not defined against key then lets just not show the icon - also look for alternate ux instead of the (i) icon
- Have to come up with some specific names for logical end point and physical endpoint - and Env can be choosen; Have a concept of default context ( we can change the env to context)
New Data store/ conn
Name of Dat Sotre:
Identify and name the first class citizens for Guzzle:
Authoring Bring data in Process and Transofmr Governance (DQ, Recon and HK) Pipelines Data Stores/Connection (logical /physical) Compute (Spark) - Straight away launches data bricks config
Monitoring and Scheduling Job monitoring
Governance New consol and dashboar standlalone dashobad.. and allow popel easily navigate DQs and Recon reluts for context param runs and datasets Launch Atlas..
/ or does it become integrated in to Atlas as additional attribute
Home page: We don't follow ADF. Rather Azure ML + Azure Portal
-
Environment show up on top banner as Subcription in Azure portal
-
Have to give me ux around how allow user to create standalone physical endpoint and link: one of the option to allow user to alternate physical conn by clicking a plus and it opens anohter tabe
-
Compute type we can reduce - data bricks.. from frontend, backend may support others..
-
Lets keep the on-prmeise features and Azure databricks feature same trunk and keep some condition logic/configs which contorls the behavour and in th workscase hvae different build script/packaging to generate speicfic build for Azure DB and On-premise
High end: Batch Job config:
- edit UI - this allows one to open all the section and use browser find..
- And new Job wizard UI both can sever diff purpose - is there a way to hve same UX to ahicev ebo
- Valdation today runs at one shot, but wiht wizard we can do at each stage
Validation: Blur on field Net step in tab Overall validation There are cross section validatoin like auto create won't work till we schema section defind as source or schema..