Having a WYSIWYG editor is a nice idea, but Summernote causes a LOT more problems than it solves. I may just ditch it entirely, but if I can find one that I can trust to not insert garbage markup, it'd be nice to have.

Ideally implemented with regular expression checking for certain keywords, but blacklisting certain IP addresses would also be handy.

Getting specific assets for image objects tends to incur extra database queries; including image_assets via prefetch/select_related whenever sunset.models.image is queried should take care of this, but a method of asset retrieval other than looping image.assets.all() would be ideal. The get_object_or_None() function in django_annoying would be ideal here; I may just go with that.

Django-processinfo was a great app, and I still like it in some ways, but it's holding back my ability to upgrade to a newer version of Django. Therefore, it's officially causing more problems than it solves.

Initially noticed this in empty categories.

I've been using Django 1.8 for way too long, and that needs to change. There are some third-party dependencies that I need to deal with before I can get this done, and probably some changes to my own code, but I can't keep putting this off any longer.

This to-do list will be modified as I explore what's needed to make this upgrade work.

• Remove django-processinfo, or modify it
• Investigate compatibility of django-static-precompiler
• Add names to django.contrib.auth views
• #68

Loading metadata is painfully slow and inefficient.

There's a lot of empty space in each link, so that can be cleaned up a bit.

Related to #32

Pingbacks aren't as popular as they used to be, but it's a feature I liked back in the days of Wordpress blogging, and there's a Django app for it. Once I get it implemented for DeerBooks (#12), I'll implement it for Sunset too.

After converting the Lupinia website to use Amazon SES for outgoing email, I now need to update the DeerConnect contact form to set the sender's email address as the Reply-To address, instead of the From address.

With the upgrade to the Showcase theme, it's now possible to implement a far better map view for DeerAttend.

Awi v3.x had support for custom HTML-esque tags, which automatically inserted a Gallery2 image block, using pipe-separated parameters. For example, <g2>25694|photo|single</g2> would be converted to <div class="one-image"><a href="http://www.lupinia.net/photo/25694.g2"><img src="http://www.lupinia.net/gallery/25696/3.g2" width="225" height="129" class="giThumbnail" alt="Assateague-1443"/> </a> </div>.

In that version of the site, the resulting markup was coming from Gallery2, and I had little control over it. The tag I defined was simply used to insert parameters into the API query to Gallery2 which returned the resulting markup. But the concept is easy enough to implement in Django.

The core bug was already fixed by switching to using an object's unicode/str method instead of the title attribute specifically, I just need to implement that fix in the breadcrumb generator.

The is_public() method has a bug that has never come up in the year or so since I implemented it, so it's quite an obscure corner case: If a model overrides that method, AND the super() call to it returns (True, ''), AND the overridden method finds a reason why something should not be public (in this case, a Sunset image that was newly uploaded and its assets had not yet been processed), the standard override template is to overwrite the value of the return of super().is_public() with new data. This does not work, because is_public() always returns a tuple.

Stack trace:

Traceback (most recent call last):
  File "/usr/local/lib/python2.7/site-packages/django/core/handlers/base.py", line 87, in get_exception_response
    response = callback(request, **param_dict)
  File "./deerfind/views.py", line 102, in not_found
    url_check = finder_function(request)
  File "./deerfind/views.py", line 48, in g2_finder
    access_check = gallery_check[0].image.can_view(request)
  File "./deertrees/models.py", line 174, in can_view
    public_check = self.is_public()
  File "./sunset/models.py", line 134, in is_public
    ispublic[0] = False
TypeError: 'tuple' object does not support item assignment

Having these separate made sense early-on, but it makes a lot less sense now that they play an increasingly-vital role in checking .can_view and .is_public. Plus, every model that extends access_control already has these fields anyway. And shifting the timestamp-related fields and methods/properties to access_control would bring some of the more advanced functionality from DeerTrees.leaf (choice of which timestamp to display) to all models.

This is mostly an issue on photos added to Sunset from Lightroom, but a method of de-duplicating tags would significantly clean up the tag list. I'm still determining how best to do this, but I'm leaning toward creating a separate model ("synonym", perhaps), and a method in the tag view for converting existing tags into synonyms of other tags.

• Build tag_synonym model
• Build admin shortcut to merge tags ("make synonym of")
• Add synonym support to deertrees.models.leaf.tag_item method
• Build 404 finder for tag synonyms
• Build admin inline view for managing tag synonyms
• Add synonyms to deertrees.views.tag_list (perhaps as a sidebar block)
• Add synonyms to keywords meta tag on leaf content pages (irrelevant, since I'm apparently not using this meta tag at all)

I'm honestly a little embarrassed by this particular bug; when I added the new filter to the subcategory widget (views.subcats) to hide empty categories, I failed to take into account the number of categories on the site that only contain other categories. Oops.

Back-end code is already in place, just needs the link.

Google does a pretty good job of crawling this site, but built-in search is crucial, and definitely something I want to add. It may require infrastructure upgrades, though; I'm not sure my server can handle ElasticSearch right now.

When .is_public()[0] is False, .is_public()[1] already contains details about why. This should be displayed.

• export_file and toc need friendly names ("Document File" and "Book", respectively)
• export_file:
• timestamp_mod in list view
  • Count of page objects in list view
• toc:
  • Help text on slug and title
  • Count of page objects in list view
• page:
  • Help text on summary, book title, book order
  • Friendly name on docfiles field

The Django admin system is awesome, and it's working just fine for me as-is, but I want to add forms that are more seamless. This would also be one way to resolve #30 without major alterations to the admin views.

body.showcase_mode .left_content .left_full.writing_page should have a max-width of 80em, to improve readability.

Similar to the DeerAttend map view, I'd love to have a view for photos that shows them on a map.

The initial blocks/layout system was a reasonable first step, and it does an ok job for most cases; different types of content (leaves) have priority ratings, and different priorities based on whether they'll be in the main content area or the sidebar. However, the number of edge-cases created by this system has significantly increased the complexity; there's a special flag indicating whether the current view is the homepage, categories and tags have to have a "content priority" setting that overrides the defaults (which can cause more problems than it solves), and I still have no particularly good way of handling special cases like /photo (which should primarily show an array of all highlighted images, not just the ones in child/descendant categories).

I'm still determining what the ideal solution would look like, but I suspect it will involve replacing the content-priority setting with an option to define custom functions/classes that the category and tag views can include. I want to keep the system flexible enough to be content-agnostic, because the "folders that can contain files of any type" paradigm is very important to me, and the DeerTrees app was created to be something that any project could potentially incorporate.

Regardless, this is in the works, and may involve a rebuild of the DeerTrees app, but it's definitely an area that's not quite where I want it to be, and I feel I can do better.

Pingbacks aren't as popular as they used to be, but it's a feature I liked back in the days of Wordpress blogging, and there's a Django app for it. I've started the implementation, but I haven't yet finished it.

Culprit is in awi_access.models.access_control.is_public(); it checks the current site. Views querysets that check this method need to include Sites via prefetch/select_related.

The database field to do this is already in place, but the template does not yet support it.

Spotted on http://www.lupinia.net/about/privacy.htm

I want to add a method for users who can edit a particular content item to be able to generate an access code for that item, which allows anyone with the code to access it for a certain amount of time. The use-case is for showing works in progress to other people (pre-readers for long-form written work, for example) without needing to publish the item or create user accounts for those people.

All system errors generate an email notification, but it would be nice to give visitors something other than "The webmaster has been notified", because that always frustrates me when websites generate an error with no way to reach out about it. But I don't want to go to the trouble of spinning up my own issue tracker, so simply using Github for that would be ideal.

My installation of Django Admin Tools is somewhat buggy, and I really like what I see in Django Jet. It's not a high priority, but it'd be nice to upgrade the site's admin views.

Goal: Reduce redundancy between main navigation links, since each one is effectively duplicated three times: Main navigation bar, mobile navigation menu, and footer navigation links.

However, I'm not quite sure whether this is viable, or whether this would solve more problems/complexity than it creates.

Newly-discovered issue that seems to only come into play when a redirect is hit by a browser with malicious information in the browser headers (something that always fails). Dealing with that is a separate issue, but the logging for pointer objects should be able to handle Unicode browser headers regardless.

DeerBooks allows pages to be written in plain text, without wrapping paragraphs in <p> tags, but the DeerTransform module doesn't interpret these paragraph/line breaks correctly, and it combines everything into one single run-on paragraph.

The deertrees/leaves.html template doesn't check whether we're looking at a category or a tag before generating these links.

Related to #5, category/tag views desperately need pagination. Which is very difficult to do with the current system without potentially dropping entire content types off of the front page, so that will probably come with the upgrade.

This is already in-progress, but not yet complete.

Error message text: "Health check failed on link 190 ({redacted}). Status code: 0".