Tinboard has turned into a tool I use pretty much every day; it's probably my most-used Textual/Python-developed application at this point. This is causing me to think more and more about how I can add things to it that are related to the core purpose, but are also outside of the main "interface with Pinboard" thing.

A thing with keeping bookmarks for a long time is that some of them go stale, go away. Some will just plain 404, others the whole site will disappear. If I find myself going back to a bookmark and seeing this is the case, I'll hit the Wayback Machine and see if there's an archive there.

So I got to thinking: what if I add the ability to perform this check into Tinboard itself? So I did just that.

Now, in the application, if you press w with a bookmark highlighted, it will check with the Wayback Machine to see if the bookmark is in the archive. If it isn't you see this:

On the other hand, if it is in the archive, you'll see something like this:

I sense this is going to be the first step in a couple of features related to this. I'm thinking that I may go on to add a "swap the URL for this bookmark with the Wayback Machine archive URL" feature, which will be handy for those bookmarks that have one away, and it would also be useful to look at the options for a "please archive a copy of this bookmark" feature.

But, for now, v0.12.0 is available and has this handy (for me anyway) first step.

Tinboard can be installed with pip or (ideally) pipx from PyPi. It can also be installed with Homebrew by tapping davep/homebrew and then installing tinboard:

$ brew tap davep/homebrew
$ brew install tinboard

The source is available on GitHub.

Back in the very early days of the Textual adventure, within the first month or so of working on the framework, we had a period of dogfooding. One of the projects I wrote during that time was a little tool I called PISpy.

The initial version was pretty much a quick hack; during that dogfooding period I did my best to try and develop a new project every couple of days. Since then I've let PISpy descend into bit rot.

In the last week or so I've turned my attention back to it and made an effort to tidy up the code, tidy it some more, and some more, and even some more.

This morning I put the finishing touches to these changes and released v0.6.0.

PISpy can be installed with pip or (ideallty) pipx from PyPI. It can also be installed with Homebrew by tapping davep/homebrew and then installing pispy:

$ brew tap davep/homebrew
$ brew install pispy

The source is available on GitHub.

While my time working on Textual might have come to an end, my time working with Textual hasn't. Three days back I experimented with Textual's newly-added "inline mode":

In doing so I extended the application so that it's possible to run tinboard add and quickly enter a new bookmark and then carry on in the terminal, without needing to "go fullscreen". I'll admit it's of limited use, but it seemed like a good shakedown of the feature and in working on it I was able to discover a couple of bugs (#4385, #4403).

The effect of this is this:

Tinboard can be installed with pip or (ideally) pipx from PyPi. It can also be installed with Homebrew by tapping davep/homebrew and then installing tinboard:

$ brew tap davep/homebrew
$ brew install tinboard

The source is available on GitHub.

Over the past year and a half I've written a lot of Python code, and a lot of that Python code has been Textual applications; most of those Textual applications have been very quick demonstration or test applications built to help support someone asking for help; some of them have been less-trivial applications written in my own time and for my own use and amusement. Of them I'd say there are two near-daily-drivers, and a couple that I either have more plans for, or like to maintain just for the hell of it.

Those latter applications are all ones that I've deployed to PyPI, and because of that are all ones that I've recommenced be installed using pipx. During that time though I've had half an inclination to make them installable via Homebrew. While probably not installable from the core Homebrew repository¹, at least installable from a "tap"² that's under my own GitHub account or something.

To this end I've had a blog post about packaging Python apps for Homebrew saved in Pinboard for a while now, and every time I look at it I think "this is a lot of faff, maybe later". Today was that "later".

As it turned out, it was way easier than I first realised. The evolution of today pretty much went like this:

Deciding to use a single repository as the "tap"

The blog post above seemed to suggest that for every application repository you want a tap for, you probably want a parallel homebrew--prefixed repository. This in turn would suggest that every time someone wants to install one of your tools, they'd need to add a new tap³. As I looked at it this seemed like way too much faff, so in the end I decided to create a single repository that I'd keep all my formula files in. The naming of homebrew-homebrew meant that the tap name would simply be davep/homebrew.

Simple and clean, I think: things for homebrew, things that can be installed via homebrew, that come from davep. To add the tap it's simply:

$ brew tap davep/homebrew

Ensuring that all my applications and libraries publish source

Although it seems that it might be (possibly, maybe, perhaps, who can tell?) deprecated, it looked like homebrew-pypi-poet was a tool I'd need to do all the heavy work on making the formula file. A quick test threw up a problem where it was complaining that my test package (one of my own applications) didn't have an sdist. Sure enough, through nothing more than never having bothered to make it happen, the source of my libraries and applications wasn't been uploaded to PyPI when I published.

So I went through some of my repositories and fixed that, making patch releases as I went.

Making a Makefile to let me be lazy

The next thing to do was to figure out the most lazy way of building the formula files. From what I could see the main steps to making all of this work were:

• Make a venv and activate it
• Install homebrew-pypi-poet
• Install the package you want to package for Homebrew
• Run poet to make the formula

Seemed simple enough. For all sorts of lazy reasons I still tend to use pipenv to get things done quickly, and that seemed to work fine here too. I'm also a fan of PIPENV_VENV_IN_PROJECT=true which makes things clean and tidy, so I figured a rule in a Makefile like this:

clean:
        rm -rf .venv
        rm -f Pipfile Pipfile.lock
        pipenv --python 3.12
        pipenv install --dev homebrew-pypi-poet

would be fine to make a clean venv ready to build the formula, and then I'd have a rule for the package itself that depended on the above, like this:

oshit: clean
        pipenv install oshit
        pipenv run poet -f oshit > Formula/oshit.rb

Fixing the package description

The above was great, and worked really well. But there was one issue that I could see: the resulting formula file always had this desc inside it:

desc "Shiny new formula"

From what I could see there was no way to tell poet what I wanted the description to be, and neither did I want to have to remember to edit that line each time I regenerated the formula file. So sed to the rescue then I guess, with this sort of thing:

sed -i '' 's/Shiny new formula/The actual text I want/' Formula/coolapp.rb

The result

The result of all of this is that I now have a repository that I or anyone else can use as a tap to be able to install my stuff using the brew command. So now if you want a little Hacker News reader for the terminal but you don't want to be messing with installing pipx and the like, but you do use brew on your machine, it's just this:

$ brew tap davep/homebrew
$ brew install oshit

Fingers crossed it all "just works" when I next upgrade one of those packages. I will, of course, have to remember to go into davep/homebrew-homebrew and make the-app for the relevant application, and then commit and push the changes, but that's really not too difficult to remember and do.

Hopefully it'll then all just work.

I just realised that it's been a while since I last posted an update about tinboard. This is probably my most-used Textual-based application, and one I'm constantly tinkering with, and just this morning I published v0.10.0.

Often the changes are small tweaks or fixes to how it works, sometimes they're simply updates to the version of Textual used, making use of some new feature or other; I've yet to add another "major" feature so far. They will come, but so far the ideas I have for the application haven't actually felt that necessary. Although I say so myself it does what I need it to do and it does it really well.

So, as a quick catch-up of what's changed since v0.4.0 (which was the last version I posted about):

• v0.5.0 was released 2024-01-04; this included all the tags of a bookmark when doing full-text searching.
• v0.6.0 was released 2024-01-10; it fixed a small bug where the tag suggestion facility got confused by trailing spaces in the input field.
• v0.7.0 was released 2024-02-02; this updated the minimum Textual version to v0.48.2 and removed all the custom changes to the Textual TextArea widget, making use of the updates to TextArea that version of Textual made available.
• v0.8.0 was released 2024-02-18; this fixed a crash on startup caused by a newer release of Textual (the fault was in tinboard; the update to Textual helped reveal the problem).
• v0.9.0 was released 2024-02-29; it simply added support for using Esc at the top level of the application to quit (I like to camp on Esc to GTFO).

Then, just now, I released v0.10.0. This release makes full use of some work I recently did to enhance Textual's CommandPalette widget, which added a "discover" system. The change in tinboard is that all of the command palette providers now have discover methods too. The result of this change is that when you open the command palette in tinboard (ctrl+p) you can see every possible command right away.

Tinboard can be installed with pip or (ideally) pipx from PyPi. The source is available on GitHub.

My association with Textual works on two levels: on the one hand, sure, it's currently my day job; on the other hand it's a FOSS project that I'm keen to support so "free time me" tries to work with it and support others working with it too. For this reason you'll often see me being terminally¹ online in the Textual Discord, trying to answer questions as they come up, every waking free moment.

Almost without exception the people who ask for help are appreciative and ask in the spirit of wanting help and wanting to work together with whoever is helping them to get an answer. That... that's actually quite a cool thing to be part of. I like the sense of community that comes from someone going "bah I'm trying to do this thing and it isn't working PLEASE HELP!".

And then... well, let's just say that sometimes the odd question will crop up that seems to be asked from a less collaborative position.

Without wanting to appear to dunk on an individual (I don't wish to), I want to break down an example that happened yesterday. For some background, I'd been AFK all day, having a wonderful time in town with a friend, shopping, lunch, a movie, that sort of thing. A nicely-chilled day where I didn't even look at the Discord notifications that had popped up on my watch and phone.

However, later on that evening, finally home and flopped on the sofa, I saw a question pop up that, while lacking any useful detail and possibly suffering a wee bit from being an XY problem, the immediate answer was clear:

BadIdentifier: 'test.udp_json_client-input' is an invalid id; identifiers must contain only letters, numbers, underscores, or hyphens, and must not begin with a number.

????????????????? a dot isn't allowed?

Like I say: it lacks context and detail, and the number of question marks doesn't really clarify much, but the core question that seems to be at play here is "is it true that a full stop can't be used as part of the ID of a widget?".

The answer is: no, it can't. There's a reason for that too, and if someone were to take a step back for a moment and think about how IDs play a part in queries and how they'd be used in a stylesheet, the reason for that might pop out. So, to help the person asking the question walk in the direction of the answer, I reply:

When you come to query that how would the parser know it’s not ID “test” combined with a class, if dot was allowed?

Before we go on, to illustrate my point, consider this ID: foo.bar. When you come to query that back, or use it in a stylesheet, how would #foo.bar look? Is it a widget with the ID foo.bar; or is it a widget with the id foo and the class bar?

As far as Textual's CSS is concerned, it would be the latter.

But at this point it didn't seem necessary to get into all that detail; I like to try and assume knowledge on the part of the person asking the question, sans any other evidence, so for the moment I'll assume a "oh, right, yeah, that's a damn good point" kind of reply. Or if not, perhaps a "I don't quite follow, could you explain?" reply, in which case I'm happy to go into all the detail.

The reply was neither:

why are element ID and classes co-mingled?

this was previously allowed so your question doesn't really make sense to me

Now I'm confused. Asking why element IDs and classes are co-mingled seems odd; but I'm used to chatting with people who don't have English as a first language so I'm going to assume it's just a wording choice; but the latter part is very odd: this has never been allowed. Or, more to the point... without any proper context I can't really appreciate what claim is being made here.

You see... I did notice a bug in Textual recently, when it came to widget IDs. Long story short: when you set your ID for the widget in your code, no actual validation of the ID was being done. This was an oversight that was fixed in the latest release.

But knowing that that's the case would be guesswork on my part; I'm also fresh at my desk after a day out; I'm probably not quite in the coding/Textual zone yet, so rather than try and guess half of the conversation, it's easier to just ask the person who is asking. So I ask them to restate the question, and give some more background.

The reply is:

the objective is to put a string like test.udp_json_client-input as the label of a tab, which previously just used the ID property. from skimming release notes, is it better to explicitly set the label, and then assign something compliant for the ID separately?

Wait... what? I thought we were talking about valid widget IDs, now we're talking about tabs and labels? Do we mean TabbeContent and the labels of a TabPane? This is a bit different. So I'm sat there trying to figure out this person's thought process so I can offer the help they're after and this follows:

okay @davep, you have a real bug. --content-tab- prefix is not ephemeral. if I create a tab and grab .id, that prefix comes with it, so if you save it for later and try to set .active, assigning .active doesn't agree that there is a tab --content-tab-thing

so I need to de-mangle the name manually before assigning .active I guess?

Wut? Like... wut? Okay, we do seem to be talking about TabbedContent, I recognise the values being mentioned here; we did some work late on last year that added some namespacing to parts of the TabbedContent widget in an effort to reduce some foot-gun situations.

But... there's no . being used in the IDs as part of that; why are we now proclaiming a bug in an unrelated PR? That's quite the leap with zero evidence. Like... sure, I'm all for being alerted to bugs and fixing them, but this doesn't seem like that.

And then there's the "so I need to..." conclusion that also seems to have no connection to the original question.

Anyone who has ever done support will recognise this situation, I'm sure. Someone has seen a problem, they've dug around a little and reached a conclusion about what the cause is, and turns up looking for help with the conclusion they've reached (very much a variant of a XY problem).

That almost never gets us where we want to go, so I do the obvious thing; I try and reboot the question; I try and get us back to the start and try and get some clarity; I try and encourage asking the question with zero assumptions:

I'm afraid I'm still not really understanding your question, as it now no longer seems to relate to what you very first asked. Perhaps you could start again, ideally with an MRE of what you're looking at and trying to do, for clarity?

I figure, whatever the problem is, it can be illustrated with like a dozen lines of code. Also, when asking people to do this, it often actually helps them rubber-duck their own problem. There's been plenty of times on Discord where someone's "found a bug" in Textual, they're asked to make an MRE of it, and they come back and go "oh, shoot, right, I did that and realised the bug was in my code". It's cool when they happens; everyone learns something.

So... no MRE comes back, but this is the reply:

I'm trying to fix multiple breakages in my application from some recent changes. Right now I can't wrap my head around what to assign a tabbed_content.active for it to work how it did before (where if you have a tab with ID sample, you can assign tabbed_content.active = "sample", but you can't do that anymore)

While not an MRE, I can work with this. It seems clear that they have a TabbedContent where they have a TabPane with the ID "sample" and they are struggling to make it the active tab by setting active to "sample". That seems hugely unlikely, this is what TabbedContent is all about, I think we'd have noticed (I'm petty sure we've got unit tests that cover this), but I'm game. I can test this. And the MRE I write will illustrate there isn't a problem.

So I reply:

Again, I can only suggest that you make an MRE of the issue you're seeing. For example, here's me making a set of tabs, the last of which has the ID "four", and I set the active to "four":

and provide the code:

from textual.app import App, ComposeResult
from textual.widgets import TabbedContent, TabPane, Label

class TabbedContentApp(App[None]):

    def compose(self) -> ComposeResult:
        with TabbedContent():
            with TabPane("One", id="one"):
                yield Label("One")
            with TabPane("Two", id="two"):
                yield Label("Two")
            with TabPane("Three", id="three"):
                yield Label("Three")
            with TabPane("Four", id="four"):
                yield Label("Four")

    def on_mount(self) -> None:
        self.query_one(TabbedContent).active = "four"

if __name__ == "__main__":
    TabbedContentApp().run()

Based on what they've most-recently said is the problem, I'm confident they'll see that this MRE is their situation in a nutshell, and we can work out from there and figure out what the problem is they're seeing and where this . in their IDs is coming from (because I'm very confident it isn't coming from the work that was done on TabbedContent).

This is good. We're getting close to heading down a good path; I can feel it!

I was wrong.

https://github.com/Textualize/textual/blob/main/src/textual/widgets/_tabbed_content.py#L513

there's no way you can deny you just added a metric ton of shenanigans with the tab ID stuff. I can't get it to work at all anymore (assigning .active), but yes I will either come up with an MRE or find the bug and let you know

So, rather than back up a wee bit, work with the MRE I wrote for them so we can take a walk through the problem, they instead decide to tell me that the PR I did last year (which still isn't implicated in any of this outwith of them seemingly assuming it's the cause of all the issues, presented with zero evidence that it is) was simply "a metric ton of shenanigans".

No!

Stop!

This is not how you ask for help.

This isn't how you ask for help from a product or service you pay for. This really isn't how you ask for help from a Free Software project, where the people who are offering you help are doing so in their free time because they want people to be able to build cool things with it.

It really isn't hard at all to show just a wee bit of respect for people's time and willingness to try and help you.

Now... I get it. I can imagine a scenario where someone has just updated Textual and their application suddenly starts throwing all sorts of weird and new errors. That happens. That happened to me on Thursday evening just gone. But that's no reason for approaching getting help like this.

The way to approach it is this: pin the problem dependency, perhaps publish a new version of your application so there's no accidental update of the dependency, then head to any of the help resources for the dependency has and work with people who want to help you to find the cause of the problem. Trust me, it'll go a lot faster if you work with them, take on board suggestions (no matter how odd they might first appear), and really don't call their code "a metric ton of shenanigans".

The conclusion to all of this? The person asking the question eventually found they were setting some widget's ID to an invalid ID; one with a . in it. So as I suspected and wanted to walk them to: they had invalid IDs all along and they only found out about this because ID validation was fixed.

Perhaps one day they'll retract the claim that my actually-unrelated code that wasn't "just" released but was from last year is "a metric ton of shenanigans". ¯\_(ツ)_/¯

I feel like I'm on a bit of a roll when it comes to building applications for the terminal at the moment; while I'm still tinkering and improving tinboard and OSHit, I had the urge to tackle another idea that's been on my TODO list for a while.

This is something I did for Emacs back in 2017 and I felt it was a perfect candidate for a Textual-based project. It's a terminal-based trivia quiz game, using the Open Trivia Database as the source of questions.

I've just published an early version to PyPI; it still needs some polish and I have a few other ideas for it, but as it stands I feel it's a fun little game to mess around with.

The idea is pretty straightforward: you can run it up and create lots of different quizzes, there are various parameters you can use to create lots of different kinds of challenges:

Once you're created a quiz, you can run it and answer away:

Right now the idea is that you answer by pressing either 1, 2, 3 or 4 (or just 1 or 2 for true/false questions); when I get a moment I'll also enable mouse support for selecting an answer too (honestly I feel keyboard-answering feels far more natural).

Once the quiz is done you can review your answers and see which were right and which were wrong:

As I say: there's a bunch of other things I want to add to this (keeping track of scores, adding session token support to reduce the chances of repeat questions, etc), but this felt like a good spot to make a v0.1.0 available if anyone else wanted to have a play.

Anyway, if this sounds like your sort of thing, it can be installed with pip or (ideally) pipx from PyPi. The source is available over on GitHub.

PS: Now you can see why I made textual-countdown.

Just a wee catch-up post about OSHit, my terminal-based HackerNews browser. Over the past couple of weeks I've made some small changes, so I thought I'd make mention of what I've done.

As of v0.5.0, which I released earlier today, I've:

• Added a quick way of following links while viewing a comment. While a comment is highlighted you can press l to follow a link; if there's more than one link in the comment a menu will be shown and you can select which one to follow.
• Added support for viewing polls. Polls seem to be few and far between on HackerNews, so when I published the first version of OSHit I didn't have one to hand to test any code against. Eventually one turned up and broke OSHit (on purpose; I wanted to see when that happened) so I could then add the code to load polls and show them. Right now it just shows scores; I might do actual charts at some point.
• Added optional item numbers in the lists; turned on/off with F4.

So far all small things, but handy little improvements. There's still a nice TODO list in the README and I will slowly work through it. Along with tinboard these are two applications that have absolutely turned into "daily drivers", so they're going to get a lot of tweaking over the next few weeks, probably even months.

This is actually the second release of OSHit since I first announced it a week back, and I'll get to that other release in a moment.

I've just published v0.2.0, which isn't a very substantial release, but which bumps the required version of Textual to v0.47.1 and has some fun with the new nested CSS feature.

Underlying the point of this release was me taking a "real world" application of mine and nesting as much of the CSS within it as possible, in part to get a feel for how and when it's useful, but also to give it a proper test in a "proper" application. In doing so I think I've found one bug.

Dogfooding is always a good idea.

The main visible change in this release is I've played around with the look of the comments dialog a bit:

I'm still narrowing this down, but I think I prefer this look to what I started out with.

Another change I made was also to the comments dialog. Before, if you performed the "expand comments" action on a comment card that already had its comments expanded, it would move focus to the first child comment; this was a deliberate choice that felt right at the time. Having used the app for a few days now I've realised that making it an open/close toggle is far more useful. So that's what I've done.

Now... as for the previous release I mentioned above. That was a fun one.

Back when I released v0.1.0 some joker decided that it would be fun to submit the blog post about it to the Orange Site. The comments there went as you'd expected:

• Some riffed off the opening paragraph, ignoring the tool itself.
• Some riffed off the opening paragraph in self-reflective way.
• Some riffed off the opening paragraph in a "I never see the problem" way.
• One or two did the usual "why even bother building that when $TOOL_OF_CHOICE exists?" dance to show their terminal purity.
• One or two posted genuinely useful links to other similar projects.
• The biggest tree of comments was kind of a fight.

One comment caught my eye though; someone reported having a problem running it. My initial thought on reading it was "my dude, seriously, you're going to report the problem in some random comment on HN rather than raise an issue with the author?!?".

For once I was wrong to be so cynical.

So, yeah, that was the reminder I needed that I'd been intentionally reckless while writing the original code, and hadn't gone back to the API code I'd written and made it behave before doing the initial release.

All of which is to say: if you run into a problem with some FOSS project, be like @mihaitodor. Issue that thing so the developer gets to know about it; don't assume they'll be reading some random comment section, social media site or Discord server!

That and don't make 500+ HTTP requests at once; that might not end well for some.

I know I'm not alone in having a relationship with the orange site that is... weird. I generally dislike the culture there, it's almost impossible to read any of the comments without being frustrated about the industry I work in or am adjacent to and some of the people who inhabit it; but as a link aggregator of stuff I might find interesting... I honestly can't think of anywhere better. So, yes, I've been a fairly avid reader of HackerNews for many years, and have even had an account there for over 4 years.

Given this, for a wee while now, I've been meaning to knock up a terminal-based client for it using Textual.

So after work on Tinboard settled down I got the urge to start a new pet project (not abandoning Tinboard, I'm still going to be tweaking and extending it of course) and finally knocking up that client seemed like the one.

Orange Site Hit is the result.

It's worth making clear from the very start: this is a read-only reader. There is no logging in, there is no voting, there is no posting of things. This is a client built with their own API and it doesn't provide such a thing; at least not now and despite me seeing past promises that this will change, there's no API for doing that sort of thing.¹

The idea of this application is you can run it up in the terminal, check the top, best and latest from the categories provided by the API, perhaps dive off into your web browser if needed, and then got on with other things.

It's there for when you're in the terminal you just need your hit of the orange site.

The main screen of the app revolves around the index of items, most of which are going to be stories. You can see an example of that above. For people who prefer things to be slightly less cramped, I've also added a "relaxed layout" mode too:

From the index you can head off into your web-browser by hitting Enter on any item; if the item is a story that links to somewhere that link will be opened; if it's something more like AskHN, or a job, it'll open the related page on HackerNews itself.

Pressing u with an item selected will let you view the details for the user who posted the item:

If you're the sort of person who wants to torture themselves by reading the comments (oh come on we all do it!), there's a comment reader/navigator too. With an item selected press c and the comment dialog will open:

I think the navigation within that dialog is fine; although I can see some scope for improvement. At the moment it uses a widget-per-comment (actually, it's at least 4 widgets per comment), which is fine and Textual handles that without an issue, even on items with lots of comments, but longer-term I can see me having some fun using the line API to build a super-efficient comment presentation and navigation widget.

That's it for now; it feels like a good v0.1.0 spot to be in. There are a bunch of things I still want to do with it (better cleaning up of the text, perhaps with some markup support so links get handled, etc; plus lots of ways of searching for stuff), but I felt it was in a place where I could start using it.

Anyway, if this sounds like your sort of thing, it can be installed with pip or (ideally) pipx from PyPi. The source is available over on GitHub.