This is mainly because i sometimes think that the tag is accepted just
because it is there in the input, but actually it is not being used at
all. I fear more people would do the same mistake.
Changed the invoice number field’s type to search to add the delete icon
on Chromium. Firefox does not add that icon, but i do not care; it is
still better that type="text".
Had to emit the change event to the numerus-tag field, otherwise the
form would not detect the change.
I also can not use keyup as a trigger because the changed modifier can
not be used in the <form>, as nothing ever changes, i do not know how to
trigger the form from children (i.e., data-hx-trigger on the <input>
does nothing), and i can not trigger for just any keyup, or i would
make the request even if they only moved the cursor with the arrow keys,
which is very confusing as Firefox resets the position (this may be due
the fact that i reload the whole <main>, but still).
In this case i have to use the same id for the dialog content in all
pages because, for now, there are a couple of forms that need to replace
it on submit—the new/edit form and the product selection form.
Unfortunately, HTMx does not have support for `formaction` attribute at
this point, so i had to use the workaround described in [0].
[0] https://github.com/bigskysoftware/htmx/issues/623
Mostly, the same problem as before: if the document title does not
reflect the application’s state, it becomes useless when there are
multiple open tabs.
In this case, however, i do not know how to tell HTMx to restore the
title to how it was before opening the dialog without a new request to
the server, that makes no sense when the dialog was closed without any
change whatsoever. Thus, i do it with JavaScript on the client side.
We do not have any design yet for errors and other notifications, so i
followed material design, for now, since we already kind of use their
input fields design.
This time i decided to use AlpineJS because there is not that much HTML
code, and the transitioning is way easier to do in AlpineJS than it
would be with plain JavaScript—not to mention the bugs i would
introduce.
This is more or less the same as a multiselect, except that now it
adds a list of string element that you write into the search element.
It is supposed to fetch a list of tag suggestions from the server, but i
have not implemented it yet.
What i really set off on was to refactor the multiselect’s x-data
context to a separate JavaScript file.
I did not see the need at first, thinking that it would not matter
because it was used only in a template and i was not duplicating the
code in my files. However, i then realized that having the context
in the template means the visitor has to download it each and every time
it accesses a form with a multiselect, even if nothing changed, and,
worse, it would download it multiple times if there were many
multiselect controls.
It makes more sense to put all that into a file that the browser would
only download and parse once, if the proper caching is set.
Once i realized that, it was a shame that AlpineJS has no way to do
the same for the HTML structure[0], for the exact same reasons: not
wanting to download many times the same extra <template> and other
markup required to build the control for JavaScript users. And then i
remembered that this is supposed to be custom element’s main selling
point.
At first i tried to create a shadow DOW to replace the <select> with
the same <div> and <ul> that i used with Alpine, but it turns out that
<select> is not one of the allowed elements that can have a shadow root
attached[0].
Therefore, i changed the custom element to extend the <div> for the
<select> and <label> instead—the same element that had the x-init
context—, but i would have to define or include all the styles inside
the shadow DOM, and bring the lang attribute, for it to look like it
did before. Out with the shadow DOM, and modify the <div>’s contents
instead.
At this point the code was so far removed from the declarative way that
AlpineJS promotes that i did not see much value on using it, except for
its reactivity. But, given that this is such a small component, at the
end decided to write it all in plain JavaScript.
It is more code, at least looking only at the code i had to write, but
i love how i only have to add an is="numerus-multiselect" attribute to
HTML for it to work.
[0]: https://github.com/alpinejs/alpine/discussions/1205
[1]: https://developer.mozilla.org/en-US/docs/Web/API/Element/attachShadow
Had to replace the tags <ul> with a div with an input, so that the
browser can focus the keywoard there. For now i do not have a
focus-within CSS rule because we do no yet have a style for focus
highlight.
I have replaced the template for-loop to fill the options with the
JavaScript equivalent for two reasons. The first is that GoLand is very
stupid and can not handle that templating code inside the JavaScript
function and complains of non-existing problemes all the time.
The second is that, taking advantage of the input, i now have filtering
of options and have to remove accents from the label and convert it to
lowercase into a separate property just for that. I could do that with
a Go function, but it is something that i also have to do for the
input’s value when it changes, therefore i am forced to use JavaScript
and, if i am already using it for one string, it makes no sense to have
duplicate functionality in Go code.
The control still has missing aria attributes, and the list of options
is not yet navigable with the keyboard.
I had in the product edit page only because it was easier to test there
while i was developing it, but it is something that should be done for
all select[multiple], of course.
I removed the whole x-cloak thing because i am not sure what would
happen if i do something wrong and Alpine can not initialize the
multiselect; probably show nothing to the user. Now it shows the
native select a fraction of a second, but if i fuck it up at least the
user can still use the app.
This was actually the (first) reason we added the tax classes: to show
them in columns on the invoice—without the class we would need a column
for each tax rate, even though they are the same tax.
The invoice design has the product total with taxes at the last column,
above the tax base, that i am not so sure about, but it seems that it
has not brought any problem whatsoever so far, so it remains as is.
Had to reduce the invoice’s font size to give more space to the table
or the columns would be right next to each other. Oriol also told me
to add more vertical spacing to the table’s footer.
Although it is possible to just print the invoice from the browser, many
people will not even try an assume that they can not create a PDF for
the invoice.
I thought of using Groff or TeX to create the PDF, but it would mean
maintaining two templates in two different systems (HTML and whatever i
would use), and would probably look very different, because i do not
know Groff or TeX that well.
I wish there was a way to tell the browser to print to PDF, and it can
be done, but only with the Chrome Protocol to a server-side running
Chrome instance. This works, but i would need a Chrome running as a
daemon.
I also wrote a Qt application that uses QWebEngine to print the PDF,
much like wkhtmltopdf, but with support for more recent HTML and CSS
standards. Unfortunately, Qt 6.4’s embedded Chromium does not follow
break-page-inside as well as WeasyPrint does.
To use WeasyPrint, at first i wanted to reach the same URL as the user,
passing the cookie to WeasyPrint so that i can access the same invoice
as the user, something that can be done with wkhtmltopdf, but WeasyPrint
does not have such option. I did it with a custom Python script, but
then i need to package and install that script, that is not that much
work, but using the Debian-provided script is even less work, and less
likely to drift when WeasyPrint changes API.
Also, it is unnecessary to do a network round-trip from Go to Python
back to Go, because i can already write the invoice HTML as is to
WeasyPrint’s stdin.
I am planning to use WeasyPrint to “generate PDF” from the same HTML
that the user view, but it seems that it does not support flex’s gap
and some other properties that i had to change to work in both user
agents.
I also moved the invoice’s “footer” inside the last product’s body
because i do not want the footer to be a “widow”.
Had to group name and description rows in tbody because i do not want
to break them on pagination.
I also could not use tfoot for subtotal, taxes, and total because then
they appear on every page.
The disclaimer should appear only at the very bottom of the last page,
but i do not know how to do that; using position fixed shows it on
every page.
They are not functions because i need to join them with the main
invoice relation, and although possible is a bit more awkward with
functions.
The taxes have their own relation because i will need them grouped by
their name in the PDF, so it will probably be a select for that
relation.
This is not yet necessary, but the empty label is because i do not want
to select a default tax for products—at least, not without a setting for
it.
Since i need to add the required attribute now to select, because
otherwise the browser would allow sending that empty value, i did not
want to do it unconditionally, just in case.
Let’s start first with a non-fancy validation method with just if
conditionals instead of bringing yet another complicated library. I
hope i do not regret it.
I wanted to move all the input field to a template because all that
gobbledygook with the .input div and repeating the label in the
placeholder was starting to annoy me. Now with error messages was even
more concerning.
I did not know whether the label should be a part of the input fields
or something that the template should do. At the end i decided that
it makes more sense to be part of the input field because in the error
messages i use that same label, thus the template does not have a say
in that, and, besides, it was just easier to write the template.
The same with the error messages: i’ve seen frameworks that have a map
with the field’s id/name to the error slice, but then it would be
a bit harder to write the template.
I added AddError functions instead of just using append inside the
validator function, and have a local variable for whether it all went
OK, because i was worried that i would leave out the `ok = false`
in some conditions.
I had started writing “constructors” functions for InputField and
SelectField, but then had to add other methods to change the required
field and who knows what else, and in the end it was easier to just
construct the field inline.
The symbols.svg files is for referencing from other SVG files with
xlink; the .glyph.json seems to be used for the search app; and the
.less file is useless to me because i do not use less.
For now i use CSS because we are not sure whether we will keep it this
way or not and, until we finally decide, with CSS is the easiest to
remove later on.
I do not have more time to update the update to the company today, but i
believe this is already a good amount of work for a commit.
The company is going to be used for row level security, as users will
only have access to the data from companies they are granted access, by
virtue of being in the company_user relation.
I did not know how add a row level security policy to the company_user
because i needed the to select on the same relation and this is not
allowed, because it would create an infinite loop.
Had to add the vat, pg_libphonenumber, and uri extensions in order to
validate VAT identification numbers, phone numbers, and URIs,
repectively. These libraries are not in Debian, but i created packages
for them all in https://dev.tandem.ws/tandem.
It works better than with the weird hover behaviour i could do in CSS,
and it already has most of the aria roles needed.
The only tricky part is to allow closing it by clicking anywhere else,
that is done by “extending” the <summary> to the whole screen, with a
lower z-index than the menu but higher than the rest of controls, that
way we force people to click on that summary.
Since users do not have access to the auth scheme, i had to add a view
that selects only the data that they can see of themselves (i.e., no
password or cookie).
I wanted to use the `request.user.id` setting that i set in
check_cookie, but this would be bad because anyone can change that
parameter and, since the view is created by the owner, could see and
*change* the values of everyone just by knowing their id. Thus, now i
use the cookie instead, because it is way harder to figure out, and if
you already have it you can just set to your browser and the user is
fucked anyway; the database can not help here.
I **am** going to use the user id in row level security policies, but
not the value coming for the setting but instaed the one in the
`user_profile`, since it already is “derived” from the cookie, that’s
why i added that column to the view.
The profile includes the language, that i do not use it yet to switch
the locale, so i had to add a relation of the available languages, for
constraint purposes. There is no NULL language, and instead i added the
“Undefined” language, with ‘und’ tag’, to represent “do not know/use
content negotiation”.
The languages in that relation are the same i used to have inside
locale.go, because there is no point on having options for languages i
do not have the translation for, so i now configure the list of
available languages user in content negotiation from that relation.
Finally, i have added all font from RemixIcon because that’s what we
used in the design and i am going to use quite a lot of them.
There is duplication in the views; i will address that in a different
commit.
jordi fita mas