Page MenuHomePhabricator

Review transaction-based approach for identifying added content
Closed, ResolvedPublicSpike

Description

T323145 produced a demo [i] for using the visual editor's existing "transaction" system for visually highlighting:

  1. What – if any – new content the software "thinks" was added during an edit and
  2. Where within an existing article the software "thinks" the new content it detects was added

This demo will help the Editing Team decide what – if any – changes we think ought to be made to how and when the visual editor draws a person's attention to the content they've just added ultimately, whether we think there is a need for us to consider an alternative technical approach to doing what "1." and "2." describe above (e.g. by using the visual diff "engine" instead, as T324736 describes).

Decisions to be made

  • 1. What – if any – changes does the Editing Team think need to be made to how and when the visual editor draws a person's attention to the content they've just added
    • The Editing Team will need to, at a minimum, prevent the edit check from thinking that the following "new content additions" warrant references:
      • Blank lines
      • Templates
      • Images
      • References themselves

The work to augment the historic to reflect the above will happen in T324730

  • 2. Will the Editing Team iterate on the approach T323145 produced to implement the changes they will have converged on in "1."? Will they instead pursue another approach (e.g. T324736)?
    • The Editing Team will continue to move forward with the approach T323145 produced.

Requirements

The software should only draw peoples' attention to content that warrants their focus so that they come to trust in – and be responsive to – the prompts the software is surfacing to them.

Where "warrants their focus" in this context means:

  • Content they explicitly added during the edit session they're currently in
  • Content the majority of Senior Contributors would agree would benefit from being referenced

Where "warrants their focus" in this context does NOT mean:

  • Content they did not add
  • Changes that Senior Contributors would agree would be confusing/not make sense to reference (blank lines, spaces, formatting changes, moving existing content around via copy and paste, etc.)

Review

CASEDESCRIPTIONEXPECTEDACTUALNOTESRESULTING ACTION(S)
1.New blank line(s)No highlight❗️HighlightNew blank lines unexpectedly highlightedNon-issue for now; revisit in T324730
2.Formatting existing contentNo highlight✅ No highlight---
3.Moving existing content (cut + paste)No highlight❗️HighlightPasted content unexpectedly highlightedKnown limitation of transaction approach
4.Moving existing content (copy + paste)No highlight❗️HighlightPasted content unexpectedly highlightedKnown limitation of transaction approach
5.Adding an imageNo highlight❗️HighlightImage unexpectedly highlightedNon-issue for now; revisit in T324730
6.Adding a templateNo highlight❗️HighlightTemplate unexpectedly highlightedNon-issue for now; revisit in T324730
7.Adding new sentence(s) (beg. existing paragraph)Highlight✅ Highlight---
8.Adding new sentence(s) (middle existing paragraph)Highlight✅ HighlightHighlight "leaked" onto a character in the sentence that follows the once I added/inserted.---
9.Adding new sentence(s) (end existing paragraph)Highlight✅ Highlight---
10.Adding a new sectionHighlight✅ Highlight---
11.Adding a new paragraph in existing sectionHighlight✅ Highlight---
12.Adding a referenceNo highlight❗️HighlightReference highlighted, along with the added sentence it was preceded byNon-issue for now; revisit in T324730
13.Splitting a paragraph with EnterNo highlight✅ No highlight------
14.Joining two paragraphs with BackspaceNo highlight✅ No highlight------

Done

  • The Editing Team has documented the revisions they think need to be made to the demo
  • Answers to all Decisions to be made are documented

i. T323145#8519309

Related Objects

Event Timeline

ppelberg renamed this task from Review to Review transaction-based approach for drawing peoples' attention to added content .Jan 12 2023, 8:03 PM
ppelberg created this task.
Esanders renamed this task from Review transaction-based approach for drawing peoples' attention to added content to Review transaction-based approach for identifying added content .Jan 16 2023, 1:49 PM

The Editing Team has documented the revisions they think need to be made to the demo

The demo was not intended to show any proposed UI for how we highlight added content (or even suggest that we do need to highlight added content), but just to show which content has been detected as added content.

The Editing Team has documented the revisions they think need to be made to the demo

The demo was not intended to show any proposed UI for how we highlight added content (or even suggest that we do need to highlight added content), but just to show which content has been detected as added content.

I think the review below hews to the guidance you shared above. Tho, @Esanders, please tell me if you think I ought to adjust how I'm reviewing the demo.

Based on what I've seen so far, it appears that cases 1., 3., 4., 5., 6. and partially 8. did not pass. Ed, assuming the below equips you with the information you need, what do you think we ought to do next?

NOTE: please comment if you can think of case that is not yet listed in the table below.

Initial review

CASEDESCRIPTIONEXPECTEDACTUALNOTES
1.New blank line(s)No highlight❗️HighlightNew blank lines unexpectedly highlighted
2.Formatting existing contentNo highlight✅ No highlight
3.Moving existing content (cut + paste)No highlight❗️HighlightPasted content unexpectedly highlighted
4.Moving existing content (copy + paste)No highlight❗️HighlightPasted content unexpectedly highlighted
5.Adding an imageNo highlight❗️HighlightImage unexpectedly highlighted
6.Adding a templateNo highlight❗️HighlightTemplate unexpectedly highlighted
7.Adding new sentence(s) (beg. existing paragraph)Highlight✅ Highlight
8.Adding new sentence(s) (middle existing paragraph)Highlight✅ HighlightHighlight "leaked" onto a character in the sentence that follows the once I added/inserted.
9.Adding new sentence(s) (end existing paragraph)Highlight✅ Highlight
10.Adding a new sectionHighlight✅ Highlight
11.Adding a new paragraph in existing sectionHighlight✅ Highlight
  1. New blank lines
  2. Adding an image
  3. Adding a template

The demo is currently highlighting everything that is inserted. At some point we will implement a heuristic to select which content needs a reference, and this will likely exclude anything that isn't a long run of text, i.e. items 1, 5 & 6.

  1. Moving existing content (cut + paste)
  2. Moving existing content (copy + paste)

These are the known limitations of this transaction-based approach. The alternative visual-diff based approach would better handle these cases.

  1. New blank lines
  2. Adding an image
  3. Adding a template

The demo is currently highlighting everything that is inserted. At some point we will implement a heuristic to select which content needs a reference, and this will likely exclude anything that isn't a long run of text, i.e. items 1, 5 & 6.

Understood. It seems like we can revisit the point you're raising here of deciding which content additions do NOT require a reference in T324730.

  1. Moving existing content (cut + paste)
  2. Moving existing content (copy + paste)

These are the known limitations of this transaction-based approach. The alternative visual-diff based approach would better handle these cases.

Understood. Before expressing an opinion about whether I think we ought to move forward with the transaction-based approach or explore the visual-diff based approach before making a decision, what – if any – edit types beyond the ones I listed and tested in T326856#8550215 do you (@Esanders) think we ought to test? Note: I'm going to pose a similar question to the team offline so that can be doubly sure we've tested this approach on the most common edit types before deciding on a path forward.

You can see the visual diff in the save dialog. If we took that approach we would use the same algorithm and basically consider anything marked up with green.

You can see the visual diff in the save dialog. If we took that approach we would use the same algorithm and basically consider anything marked up with green.

Understood.

Alright, @Esanders, I've finished reviewing – what I understand to be – the last three cases to be tested: 12., 13., and 14.

Before closing this task, can you please confirm whether you agree with the RESULTING ACTION I named for 12.: the added reference being highlighted right now is not an issue and we can specify that added references be excluded T324730?

You can see the visual diff in the save dialog. If we took that approach we would use the same algorithm and basically consider anything marked up with green.

Understood.

Alright, @Esanders, I've finished reviewing – what I understand to be – the last three cases to be tested: 12., 13., and 14.

Before closing this task, can you please confirm whether you agree with the RESULTING ACTION I named for 12.: the added reference being highlighted right now is not an issue and we can specify that added references be excluded T324730?

With:

  1. Ed having confirmed the above during today's team meeting
  2. Me updating T324730 to include the four cases (blank lines, images, templates, and references) we identified in this review
  3. Me updated the task description to document the answers to the 2 "decisions to be made"

...this task can be resolved

ppelberg updated the task description. (Show Details)