From 43f7b4ba9c055f87b9f446f269e12d5e11d0437f Mon Sep 17 00:00:00 2001 From: Adam Wright Date: Tue, 22 Sep 2026 04:12:23 +0000 Subject: [PATCH 1/2] Answer the curators' 19 September round, and track what is left They had added a round of findings to the report document that nothing here had answered, marked in purple for "resurfaced or still an issue on recheck". Read via the snapshot script; the document itself is theirs and untouched. Answered in CURATOR-REPORT.md, and the four that are real are now issues rather than lines in a Google Doc: - #296 the hit reaction count is one off, moving FDR and the entity ratios. The most serious of the round: everything else is something being hard to see, this is a number being wrong, and it propagates into two more. - #297 hovering a subpathway in the hierarchy does not highlight it on an EHLD, which the live site does. - #298 H2O and H+ disappear when zooming out, contradicting an OK on that row. The contradiction was in a comment, which is where such things get lost. - #299 PPTX entities move but an icon's border and centre are separate shapes, as is each reaction line segment. Grouping, not fewer shapes. And #154 is corrected rather than closed: #171 made the *table* readable, and the request -- which components of a complex are hit, shown on the diagram -- is still outstanding. They found the workaround themselves, which is not the same as it being fixed. **Right-click is not broken.** Two reported no menu, one adding "I use the touch pad though", and that is it: driven on beta, a right-click on an entity opens a 354x213 panel over the diagram with the entity's name and its three entries. A trackpad sends no right-click at all unless secondary click is enabled, so the report is about input rather than about the diagram. The answer says how to turn it on, and asks them to come back if Ctrl-click also does nothing, because then it is ours. Also answered: the four questions they left as comments, including two we cannot answer with an explanation because the honest answer is that the panel is unclear and that is our bug. Co-Authored-By: Claude Opus 5 --- CURATOR-REPORT.md | 63 +++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 63 insertions(+) diff --git a/CURATOR-REPORT.md b/CURATOR-REPORT.md index 3573d12b..2c9f382d 100644 --- a/CURATOR-REPORT.md +++ b/CURATOR-REPORT.md @@ -14,6 +14,69 @@ saves us guessing. Last updated: 2026-09-20 +## Answers to the 19 September round + +Everything you marked in purple on the report, answered. Where I could check it +I did, and I say which; where I could not, I say that instead of guessing. + +### Right-click in the diagram — it works, and it is your trackpad + +Two of you reported no right-click menu, one noting "I use the touch pad +though". That is almost certainly it. Driven on beta against +`R-HSA-1368108`, a right-click on an entity opens the panel: 354x213 pixels, +positioned over the diagram, carrying the entity name and the Molecule / +Pathways / Interactors entries. + +A trackpad sends no right-click at all unless secondary click is switched on. +On a Mac that is **System Settings -> Trackpad -> Secondary click**, or hold +**Ctrl** and click. On Windows, a two-finger tap. + +If it still does nothing with Ctrl held, tell us, because then it is ours. + +### Still open, and we agree with you + +- **Which components of a complex or set are hit** — you said on 19 Sept you + still cannot see this in the diagram. #154 is open. The analysis _table_ can + flag each hit, which is the workaround you found; showing it on the complex + itself is the request ([#154](https://github.com/reactome/WebsiteAngular/issues/154)). +- **Hit reaction count is one off**, which moves FDR and the entity ratios + ([#296](https://github.com/reactome/WebsiteAngular/issues/296)). The most + serious thing in the round, because every other finding is something being + hard to see and this one is a number being wrong. It needs the slide you + referenced, or any analysis token and the pathway where the count differs. +- **PPTX granularity** — entities move now, but an icon's border and centre are + separate shapes, and each reaction line segment moves alone. Partially fixed + is a fair description ([#299](https://github.com/reactome/WebsiteAngular/issues/299)) — the fix is grouping each entity's parts, not fewer shapes. +- **EHLD hover highlight** — hovering a subpathway in the hierarchy highlights + it on the live site and not here ([#297](https://github.com/reactome/WebsiteAngular/issues/297)). +- **Trivial molecules at zoom** — H2O and H+ still disappear when zooming out, + which contradicts the OK on that row ([#298](https://github.com/reactome/WebsiteAngular/issues/298)). We will take yours. + +### Questions you asked + +- _"Do you mean search for something in the diagram?"_ — yes. Type in the + diagram's search box, then flag the result. +- _"How do you opt to download the figure rather than the EHLD?"_ — a fair + question and we do not think the panel makes it clear. Treating it as a bug + in the panel rather than something to explain. +- _"Will we know a GIF or PPTX failed, or must we open it?"_ — you will know: + a failure produces no file and an error, not a broken one. If a file arrives, + it was produced. +- _"I did not see an option to convert to editable shapes"_ — there is none to + find. PPTX shapes are already editable when the file opens; nothing needs + switching on. + +### Fixed since you last looked + +- **A searched entity being deselected when you navigate within a diagram** + (#168). Fixed. `select` was being overwritten with the pathway you were + leaving, so your own selection was discarded on one of the two ways out of a + diagram and kept on the other. +- **An error thrown when deleting a custom interactor resource** (#231). +- If you write a test URL by hand, the flag parameter is `flag=`. `FLG=` also + works. `flg=` is silently ignored and nothing will be flagged, which is worth + knowing before concluding flagging is broken. + ## Please check on beta.reactome.org > Note the URL. `beta.reactome.org` is the new site. `reactome.org/beta` is an From 0043055333820f79761d802fc326e2bf922d602d Mon Sep 17 00:00:00 2001 From: Adam Wright Date: Tue, 22 Sep 2026 04:31:42 +0000 Subject: [PATCH 2/2] Correct what we told curators about failed downloads Found reviewing this before merging it, which is the point of doing that separately: the answer was confidently wrong about three of the five formats. I had written that a failed download produces an error and no file, so a file that arrives was produced. That is true of **GIF and PPTX** -- the two they asked about -- because those come from the render service, which builds the whole file before it answers, and `FileDownloadService` saves nothing unless the whole of it arrived and the server said it was good. It is not true of **SBML, SBGN and PDF**. Those come from the Java exporters, which stream while they generate, so a failure partway produces a truncated file that looks complete and carries no length to check against. `file-download.service.ts` says exactly this at the top, and I wrote the answer without reading it. The half I got wrong is the worse half to get wrong: a curator told "a file that arrives was produced" would stop opening the ones that can lie. Corrected, and the earlier claim is named rather than quietly replaced, because they may already have read it. Co-Authored-By: Claude Opus 5 --- CURATOR-REPORT.md | 19 ++++++++++++++++--- 1 file changed, 16 insertions(+), 3 deletions(-) diff --git a/CURATOR-REPORT.md b/CURATOR-REPORT.md index 2c9f382d..ddeee210 100644 --- a/CURATOR-REPORT.md +++ b/CURATOR-REPORT.md @@ -59,9 +59,22 @@ If it still does nothing with Ctrl held, tell us, because then it is ours. - _"How do you opt to download the figure rather than the EHLD?"_ — a fair question and we do not think the panel makes it clear. Treating it as a bug in the panel rather than something to explain. -- _"Will we know a GIF or PPTX failed, or must we open it?"_ — you will know: - a failure produces no file and an error, not a broken one. If a file arrives, - it was produced. +- _"Will we know a GIF or PPTX failed, or must we open it?"_ — for **GIF and + PPTX**, you will know: those come from the render service, which builds the + whole file before it answers, and the page saves nothing unless the whole of + it arrived. A failure is an error on screen and no file. + + **Not true of every format**, and the difference is worth knowing. **SBML, + SBGN and PDF** come from the Java exporters, which stream while they + generate: a failure partway produces a **truncated file that looks + complete**, with no size known in advance to check against. Those are the + ones to open and verify. + + An earlier draft of this answer said "if a file arrives, it was produced", + full stop. That is right for the two you asked about and wrong for the other + three, which is the worse half to get wrong — so it is corrected rather than + left for somebody to trust. + - _"I did not see an option to convert to editable shapes"_ — there is none to find. PPTX shapes are already editable when the file opens; nothing needs switching on.