Feb 012017

I coded XML.com in Wag­tail, a CMS based on Djan­go. It works well for my needs and I like Python as a pro­gram­ming lan­guage. One of the big reas­ons I like Wag­tail is that it includes a power­ful enough but not overly com­plic­ated work­flow with roles and a built-in mod­er­a­tion and pre­view sys­tem.

But, I wanted a sys­tem where people could sub­mit news items that would go into the mod­er­a­tion queue without need­ing to sign up for a login first. For­tu­nately, Wag­tail makes that pos­sible, and there’s a nice art­icle by Erin Mul­laney at Wag­tail: 2 Steps for Adding Pages Out­side of the CMS that details all the steps you need. It all worked nicely in more recent ver­sions of Wag­tail (thanks, Erin!) except for one part, the noti­fic­a­tion that the news item is in the mod­er­a­tion queue. That wasn’t a stop-ship item, so XML.com launched without those emails work­ing.

I’ve now found the source of the prob­lem. It turns out that when you sub­mit a news item in this way, it doesn’t have a login iden­tity attached to it (obvi­ously, since there isn’t one). The send_notification func­tion that sends the email uses tem­plates, and these tem­plates use the login iden­tity of the author in the body of the email. Since that doesn’t exist, the whole func­tion fails. 

That means the solu­tion is easy. The affected tem­plates are wagtailadmin/notifications/submitted.txt and wagtailadmin/notifications/submitted.html, and Wag­tail lets you cus­tom­ize the admin tem­plates. I put my cus­tom­ized admin tem­plates into a utils applic­a­tion, which con­tains all my util­it­ies for the site. My utils/templates/wagtailadmin/notifications/submitted.txt file now has the con­tent

{% extends 'wagtailadmin/notifications/submitted.txt' %}
{% load i18n %}

{% block content %}
{% blocktrans with page=revision.page|safe %}The page "{{ page }}" has been submitted for moderation.{% endblocktrans %}

{% trans "You can preview the page here:" %} {{ settings.BASE_URL }}{% url 'wagtailadmin_pages:preview_for_moderation' revision.id %}
{% trans "You can edit the page here:" %} {{ settings.BASE_URL }}{% url 'wagtailadmin_pages:edit' revision.page.id %}
{% endblock %}

Sim­il­ar changes are neces­sary for the wagtailadmin/notifications/submitted.html file if you want to send HTML emails instead.

May 102016

If you read the doc­u­ment­a­tion closely enough, of course all the inform­a­tion is there. Get­ting the order of oper­a­tions right, how­ever, can cause the odd issue.

Devel­op­ing Djan­go apps means apply­ing migra­tions, and those don’t always do what’s expec­ted. In that case, you can roll back to the n-1 migra­tion by using ./manage.py migrate [app_label] {n-1_migration_label}, then delete the nth migra­tion, then edit the models.py to try again.

To clean up the data­base from some third-party app you decide you don’t want after all, you use ./manage.py migrate [app_label] zero to get rid of the migra­tions from that app. You have to run this before delet­ing the app from your settings.py file.

Apr 242016

One of my cli­ent web­sites sud­denly star­ted giv­ing an error: Error estab­lish­ing a data­base con­nec­tion. When I went to the /wp-admin URL, the error was still there.

This par­tic­u­lar web­site is on shared host­ing, so I logged into the CPan­el and checked the data­base was still there. Then I checked the data­base and found some issues with some of the tables.

[site.wp_links] error: Table upgrade required. Please do "REPAIR TABLE `wp_links`" or dump/reload to fix it!
[site.wp_options] error: Table upgrade required. Please do "REPAIR TABLE `wp_options`" or dump/reload to fix it!
[site.wp_postmeta] status: OK
[site.wp_posts] status: OK
[site.wp_term_relationships] status: OK
[site.wp_term_taxonomy] error: Table upgrade required. Please do "REPAIR TABLE `wp_term_taxonomy`" or dump/reload to fix it!
[site.wp_terms] status: OK
[site.wp_usermeta] error: Table upgrade required. Please do "REPAIR TABLE `wp_usermeta`" or dump/reload to fix it!
[site.wp_users] error: Table upgrade required. Please do "REPAIR TABLE `wp_users`" or dump/reload to fix it!

Run­ning those SQL quer­ies on the appro­pri­ate data­base in phpMy­Ad­min fixed the prob­lem. I don’t know wheth­er the host­ing com­pany upgraded the data­base, or some­thing happened with the auto­mat­ic Word­Press upgrade sys­tem, or if some­thing else caused the prob­lem.

[Update] There were a bunch of oth­er errors that cropped up after­wards with the White Screen of Death; I had to call the host­ing com­pany to sort out the server-side errors caus­ing those. It’s pos­sible those errors were the ori­gin­al cause of the data­base prob­lems, whatever they were.

Apr 222016

I dis­covered another issue while deploy­ing to Python­Any­where (may­be it’s applic­able to oth­er PAAS pro­viders as well).

There was an odd Impor­t­Er­ror when run­ning manage.py. In the spe­cific case I had, it showed up when run­ning the tests with coverage: from Uni­path import Path Impor­t­Er­ror: No mod­ule named ‘Uni­path’. It turned out I hadn’t installed cov­er­age in the vir­tu­al envir­on­ment, which meant the sys­tem was using the default one. Installing cov­er­age in the vir­tu­al envir­on­ment as well fixed the prob­lem.

Apr 212016

A check­list for mov­ing a Django-Wagtail pro­ject to Python­Any­where. There is doc­u­ment­a­tion on the Python­Any­where site; mine includes things I for­get.

Setup: devel­op­ment and test­ing on my laptop, sta­ging and pro­duc­tion on Python­Any­where.

The help files are pretty good, but I need my own check­list. Right now I’m in the sta­ging mode, but at some stage I’ll be mov­ing to pro­duc­tion. No point fig­ur­ing out the same things twice!

  1. Develop on laptop in a vir­tualenv. Push com­mits reg­u­larly to bit­buck­et account. At some stage squash the migra­tions and clean those up. Four sets of set­tings: dev, test­ing, sta­ging, pro­duc­tion.
  2. Set up account on Python­Any­where that allows the use of Post­gres (it’s an add-on to a cus­tom plan).
  3. Cre­ate vir­tualenv and set up sta­ging web app. Delete vir­tualenv when you real­ise you didn’t use the right ver­sion of Python and the default is 2.7, not 3.5. Recre­ate the vir­tualenv with python 3.5.
  4. Clone the repos­it­ory (using the ssh-keygen instruc­tions). Redir­ect the pub­lic key to a file so you can copy it without line-breaks get­ting in the way.
  5. pip install -r requirements/production.txt (includ­ing psycopg2, which I didn’t need for devel­op­ment).
  6. Cre­ate the Post­gres server, user, and data­base Don’t for­get a strong pass­word for the user (own­er of the pro­ject data­base).
  7. Update the set­tings file with the data­base set­tings.
  8. Set the envir­on­ment vari­ables for the set­tings and the secret key (gen­er­at­or).
  9. Attempt to apply the migra­tions. This will show where you made mis­takes on all the pre­ced­ing steps.
  10. Fix the mis­takes. Reload the web app to see if any­thing shows up. 
  11. Set up the stat­ic file server. Check the stat­ic files are being served cor­rectly.
  12. Cre­ate the Djan­go super­user and log in.

The next step is data, of course.

Jan 022016

Over the Christ­mas break I made a couple of dips, one of which got bet­ter reviews than the oth­ers. This is not a recipe for pur­ists, since a real tapen­ade should have anchovies in it, but I didn’t have any and my fam­ily doesn’t like them any­way.

None of the quant­it­ies are exact. The sun-dried toma­toes were loosely packed in the meas­ur­ing cup and I didn’t meas­ure the olives, just drained the can and tossed them in the food pro­cessor. I didn’t chop any­thing before put­ting it in the food pro­cessor.

  • Approx 2 cups black olives (con­tents of one can, 398ml size). I used Cali­for­ni­an black olives since those were in the cup­board, next time I’ll prob­ably use Kala­mata olives.
  • Approx 3/4 cup oil-packed sun-dried toma­toes; let most of the oil drip off but not all of it.
  • 5 cloves of gar­lic.
  • 2 tbsp capers

Pro­cess in a food pro­cessor until finely chopped. If it’s too dry, add a few drops of olive oil (or oil from the sun-dried toma­toes).