-
Notifications
You must be signed in to change notification settings - Fork 24
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
My thoughts on changes to this. #415
Comments
There's definitely a lot there we could improve on. And finding a way to developer more quickly would be good. I need to get to sleep, but wanted to say that this is on my queue to reply in more detail. |
Here's an update.
|
Thanks being able to see the effects of the PRs is great for faster reviews.
Lets follow up on that one in the other thread. There's a level that we need to make the standardization and decisions at the rosdistro maintenance level before we put it all into documentation.
This is great! I actually just filed #425 and #426 before getting to this issue because it was bugging me so much when I was testing the site for the JS changes. The speed is great! The data density is much better. Avoiding the duplication is great. We don't need the cross-distro work with the tabs. In my other issue I think that adding the number of rows available will help people know to look in other distros if they are not getting what they want.
Yeah, if it's fast enough to reload that's fine. And if the json is static it should get cached by the browser and or CDN/proxies.
I"m not as worried about the space above. For me the problem is that the bottom doesn't flow down. Also in usability, the right scroll bar doesn't drag properly for me. I could scroll wheel and page up and down, but not drag the scrollbar to jump to the top. And the distro column is now redundant with the tab filter at the top so could make more space.
The existing tags are mostly I think for rendering flags in the package display versus for user consumption. But potentially exposing some of them might make sense. We could then add filters to the search as well based on the tags. |
Yes, this is far from the final form. It is mainly meant to be a trial of Tabulator to see if this is a viable approach. So far I am impressed. These development builds will have lots of glitches for awhile. The height of the visible rows is a parameter that is set when creating the table. I could show a lot more rows, but there are advantages to being able to see all displayed rows on a single desktop computer screen. Not sure what the right tradeoff is here. But the fixed height is needed for the speed, as Tabulator uses a virtual DOM and only displays the rows that are visible. You could add some logic to adjust the height depending on the size of the available screen, but it is never going to respond smoothly to that like it would if you could rely on the browser to resize in CSS. Right scroll works fine for me, not sure about your issues. Tabulator is a large, complex project so we might have to live with whatever limitations or bugs it has if we use it. I'm currently working on merging a filtering functionality into the display, then I will also add "filtering" via the Lunr search, to have the unified display I mentioned above. Yes tags and distro are currently useless on this screen. But tags I believe are an important feature long-term. I keep distro partly because distro switch is buggy with this, and until there is a clear decision to limit this display to a single distro. |
If we have a filterable display that is quick, then the package list can just be the search without any search string. And the filtering for tags or distro are also just more advanced search functions too. So we can really collapse those capabilities fully potentially. I think that this experience is better than our current pagination approach so lets see where it can go. |
I integrated search into https://devindex.rosdabbler.com It's amazing how much more effective search is with just searching over a single distro. That really narrows the search, and gets rid of the problem of zillions of tf2 results, all in different distros. I was expecting to have to deal with several hundred entries per search, but it seems like I am getting several tens at most. Those are not difficult to visually search with the scrollable list. So I am questioning the value of the filtering separate from search that I was planning to add. Perhaps a simple text search is going to be enough? So far, this has been a conceptual demo. There are many UI glitches, but the search functionality seems to be working well, and a major improvement over the current experience. I think I will work on solving some of the UI issues (like getting the proper table height) and see if we can land this. |
Yeah, that search is definitely working well. Snappy and easy to get to what you want already. I think that the filtering is a nice to have, but it's not at the top of my priority list. For the top bar I know that you were looking to compress it. And I agree with making the package listing, repo listing, and rosdep listings always visible. But maybe it would look less compressed to keep them in the blue menu bar at the top (where they used to be a drop down) instead of the larger clickable boxes that take up a lot of space. |
The design for the package listing, repo listing, and rosdep listings was copied from the previous version. I also find that ugly. But they need to stand out somehow from the external links which currently are the upper page items. After playing some with Tabulator layout, I think that the overall design of the web page needs to change to properly show it responsively. Basically Tabulator needs a fixed height web design to properly enable its virtual dom (or at least a preset height on the table), which I believe is important for proper performance. If the design is going to change completely, it might be time to also do two other changes at the same time:
I'm going to experiment with doing that just on the package list page initially rather than tackle the whole site. Not sure if that will be feasible or not yet. Also it would be time to think through not only responsiveness (effect of screen size) but also accessibility. I'm not really a web designer, so I am learning as I go how to do this. If there are people more suitable for the task feel free to recommend yourself or others, but barring that I will muddle through somehow. |
I'm very much muddling through web design too. Upgrading bootstrap is fine. And JQuery use or not is an implementation detail not an architectural decision. Though it's a pretty low level tool that is usually pretty useful. Focusing the testing in one place sounds like a good idea. I look forward to seeing what you can do with Tabulator. |
Update:
|
Too bad about the bootstrap changes. I agree with not trying to do the bigger chagnes. The new layout looks reasonable. One thing I noticed is that it took ~40 seconds for me to load the package index content when I did a force reload one of the times. Maybe that was the dev hosting spinning up. As I can't reproduce it now. |
Latest update, available at https://devindex.rosdabbler.com/, was focused on getting the package list page to only scroll in the Tabulator section. This now WFM on desktop. On mobile (FF on a Pixel 6) it scrolls in the browser for the height of the URL bar. I believe though this is a feature of the mobile browser, so I am not concerned. Earlier @tfoote reported scrolling problems, hopefully those are now gone. The next step is to change what fields are actually shown in the listing, searched, and filtered (should that make sense to do.) Concerning "it took ~40 seconds for me to load the package index " there was a build running on the machine at the time. The web server is currently serving directly from the _site directory, so during builds it might not only be slow, but non-functional. Usually they only run in the night, but sometimes I trigger them manually when I am working on stuff. |
The site at https://devindex.rosdabbler.com/ is now fully completed. Last chance for comments before I start submitting PRs to land this. Major changes since the last comment here:
I am not aware of any issues, so please report any glitches. In the version presented, the old "Search" pages can still be accessed, but the only way to reach them is from a search bar on the home page. My expectation is this will be completely removed along with the old packages list code when the system deps page is made to be searchable. There is no direct link now to the old packages list, but it can still be accessed for now with /packages (though the intention is to remove it completely). The full landing can be done in a few PRs, roughly:
|
In case anyone follows this, there was an important experimental update today. I created a repo https://github.com/rkent/ros_webdata that uses github actions to keep updated a few items that are slow to update, but also slow to change. This includes 1) descriptions for pip dependencies, and 2) package download counts from osuosl.org. The intention is that this repo would be transferred to some ros organization if this idea has merit. https://devindex.rosdabbler.com now include both of these, that is pip descriptions as well as package download counts. I have no doubt that pip descriptions will eventually be added in some form or another. The package download counts could be somewhat more controversial, but I find it useful to know what are the popular ROS packages. |
I've been thinking for awhile about proposing some major changes to rosindex. Here are some not-fully-developed thoughts that I have had. I would appreciate any comments from others on directions.
I've always found rosindex hard to use for a variety of reasons. Some general issues:
The front page has all of these links outside of rosindex. Once you start using it, the core functionality (showing packages, repos, and system dependencies) is behind a drop-down menu, while plenty of external or less-important items are one click.
The "packages" page is 39 pages long, with no filtering available to narrow the focus.
I don't know what the stated purpose of rosindex is, but it seems to me the main function should be package discovery. A secondary purpose today is also clearly to give package details.
What would I change?
This can be done incrementally from the present site, but my experience with rosdoc2 is that this is going to be review limited. But it might be worth considering having a development site with looser review rules, planning to switch to a revised rosindex as a major update rather than incrementally.
Thoughts?
The text was updated successfully, but these errors were encountered: