How to fix "StackLevelError (Stack Overflow)" in Jekyll navigation template - ruby

I am trying to write a recursive Jekyll navigation template (include) as described in "Nested tree navigation with recursion". I have a minimal example committed in jekyll-min, which basically has:
two top-level dirs, each with one page
another dir under the second top-level dir, containing one page
a navigation template (_includes/docs_contents.html) that loops through the top-level dirs and initiates recursive traversal for each
a recursive include (_includes/nav.html) that accepts a navigation entry, renders its title and child links, and invokes itself recursively for any dirs in its children list
a layout (_layouts/doc.html) that renders the navigation pane and content for each page
I'm using Ruby v2.7.0 and Jekyll v3.8.5.
# docs structure
_docs
|
|_a/
| |_index.md
|
|_b/
|_index.md
|
|_1/
|_index.md
# _data/docs-nav.yml
- title: a
docs:
- link: /a/
- title: b
docs:
- link: /b/
- title: 1
docs:
- link: /b/1/
# _includes/nav.html
{% assign section=include.nav %}
<div class="ui accordion">
<div class="title active">
<i class="dropdown icon"></i>
{{ section.title }}
</div>
<div class="content active">
<div class="ui vertical text menu">
{% for item in section.docs %}
{% if item.link %}
{%- assign p = site.documents | where: "url", item.link | first %}
<a {%- if page.url== p.url %} class="current item" {% endif %} class="item" href="{{ p.url }}">
{{ p.menu_name | default: p.title }}
</a>
{% endif %}
{% if item.docs %}
{% include nav.html nav=item %}
{% endif %}
{% endfor %}
</div>
</div>
</div>
# _includes/docs_contents.html
<div class="unit one-fifth hide-on-mobiles">
<aside>
{% for section in site.data.docs_nav %}
{% include nav.html nav=section %}
{% endfor %}
</aside>
</div>
# _layouts/doc.html
---
title: Docs
description: version 1.0
---
<html>
<body>
{% include docs_contents.html %}
{{ content }}
</body>
</html>
As far as I understand, for each page the navigation template render should work like this:
_layouts/doc.html
_includes/docs_contents.html: iterate root level entries, calling _nav for each
_nav(/a/ entry): render title, iterate docs, render /a/ link, and quit
_nav(/b/ entry): render title, iterate docs, render /b/ link, and then call _nav(/b/1/ entry)
_nav(/b/1/ entry): render title, iterate docs, render /b/1/ link, and quit
_nav(/b/ entry) (already in stack): quit
_includes/docs_contents.html: quit
However, when I perform a bundle exec jekyll build I get:
Liquid Exception: Liquid error (/mnt/e/ThirdParty/jekyll-min/_includes/docs_contents.html line 17):
Nesting too deep included in /_layouts/doc.html
jekyll 3.8.5 | Error: Liquid error (/mnt/e/ThirdParty/jekyll-min/_includes/docs_contents.html line 17):
Nesting too deep included
Traceback (most recent call last):
[...]
What is the problem with my content or the recursive template? I have been struggling with this for hours with no luck.
JEKYLL_LOG_LEVEL=debug
didn't produce any additional useful info.
The actual document structure is more complex and could go arbitrarily deep, so writing a non-recursive template to manually handle nested levels may not be an option.

Excellent question.
With help of {{ myvar | inspect }} and a flag limiting recursion, I've successfully debugged your code and understood why this infinite recursion occurs.
It comes from the fact that the section variable in docs_contents.html is assigned by in a for loop and freezed : it cannot be changed.
The first time you include nav.html, {% assign section=include.nav %} is not changing section and your code just use the one assigned in your for loop.
When you recurse and call nav.html a second time it will use the same freezed global section variable and recurse indefinitely.
The solution is to change your variable name in nav.html from section to something else. eg: sub_section, and it will work, because this new variable will not be freezed and can be reassigned as needed during recursion.
{% assign sub_section=include.nav %}
{{ sub_section.title }}
{% for item in sub_section.docs %}
...
If you want to experiment here is my test code with some comments :
docs_contents.html
{% for section in site.data.docs_nav %}
{% comment %} ++++ Try to reassign "section" ++++ {% endcomment %}
{% assign section = "yolo from docs_contents.html" %}
{% assign recursion = 0 %}
<pre>
>> docs_contents.html
++++ "recursion" var is assigned and becomes global
recursion : {{ recursion | inspect }}
++++ "section" is freezed to loop value ++++
including nav with include nav.html nav=section >> {{ section | inspect }}
</pre>
{% include nav.html nav=section %}
{% endfor %}
nav.html
{% comment %} ++++ Try to reassign "section" ++++ {% endcomment %}
{% assign section = "yolo from nav.html" %}
<pre>
>> nav.hml
recursion : {{ recursion }}
include.nav : {{ include.nav | inspect }}
++++ "section" is freezed to loop value ++++
section : {{ section | inspect }}
</pre>
{% comment %} ++++ useless assignement ++++ {% endcomment %}
{% assign section=include.nav %}
{% for item in section.docs %}
{% if item.link %}
{%- assign p = site.documents | where: "url", item.link | first %}
<a {%- if page.url== p.url %} class="current item" {% endif %} class="item" href="{{ p.url }}">
{{ p.menu_name | default: p.title }}
</a>
{% endif %}
{% comment %}++++ limiting recursion to 2 levels ++++{% endcomment %}
{% if item.docs and recursion < 2 %}
{% comment %}++++ incrementing "recursion" global variable ++++{% endcomment %}
{% assign recursion = recursion | plus: 1 %}
{% include nav.html nav=item %}
{% endif %}
{% endfor %}

Related

How to iterate through three lists in parallel in Yaml file

Need 3 variables to change on every iteration.
ver = [1,2,3]
tes = [a,b,c]
bet = [1a.2b.3c]
{% for v,t,b in ver,tes,bet %}
{{ v }} {{ t }} {{ b}}
{% endfor %}
o/p:
1 a 1a
2 b 2b
3 c 3c
I ran the same, got a very messy o/p. There should be only 3 iterations printing 3 variables in each iteration.
How can i perform the above operations using the for loop ? In the above case
Any suggestions on where I'm going wrong, would help Thanks !
You can use the following code:
{% for i in range([ver|count, tes|count, bet|count]|min) %}
{{ ver[i] }} {{ tes[i] }} {{ bet[i] }}
{% endfor %}
It uses only built in filters and functions:
count
min (to handle sequences with different length, similarly as zip in python)
range
Entire code in python:
import jinja2
ver = [1, 2, 3, 4]
tes = ['a', 'b', 'c']
bet = ['1a', '2b', '3c']
t = jinja2.Template('''{% for i in range([ver|count, tes|count, bet|count]|min) %}
{{ ver[i] }} {{ tes[i] }} {{ bet[i] }}
{% endfor %}
''')
print(t.render(ver=ver, tes=tes, bet=bet))
It looks like you need something that behaves like Python's zip function. Here's one option, in which we simply expose zip to the Jinja environment:
>>> import jinja2
>>> ver=[1,2,3]
>>> tes=['a','b','c']
>>> bet=['1a','2b','3c']
>>> t=jinja2.Template('''{% for v,t,b in zip(ver,tes,bet) %}
... {{ v }} {{ t }} {{ b}}
... {% endfor %}
... ''')
>>> t.environment.globals.update(zip=zip)
>>> print(t.render(ver=ver,tes=tes,bet=bet))
1 2 3
a b c
1a 2b 3c
>>>

Round off the values in a list in jinja

I have a dictionary as
dictionary_num = {'Age': [-1.0, 25.0], 'M2': [-1.0, 42.0], 'younChild': [-1.0, 7.0], 'M4': [-1.0, 216.0]}
and another dictionary as
dictionary_mcqsr = {'M5': [-1.0, 3.0, 0.0, 2.0, 1.0], 'M6': [-1.0, 0.0, 1.0], 'M7': [-1.0, 1.0, 0.0], 'M8A': [0.0, -1.0, 1.0, 2.0], 'M8B': [1.0, 0.0, -1.0], 'M9': [0, 1], 'M11': [0, 1], 'M12': [2, 1, 0, -1]}
which is rendered as
{% for key,value in dictionary_num.items() %}
{% if key == j %}
<td>{{value}}</td>
{% endif %}
{% endfor %}
{% for key,value in dictionary_mcqsr.items() %}
{% if key == j %}
<td>{{value}}</td>
{% endif %}
{% endfor %}
where j refers to the keys. When I render it, I want to remove the decimal points at the jinja side in both dictionary_num and dictionary_mcqsr and sort the values for dictionary_mcqsr but could not find any solution.
Use a filter around what you want to round off :
Example : (for a single element)
{{ (value)|round|int }}
EDIT : For the list :
{% for key,value in dictionary_num.items() %}
{% if key == j %}
<td>[{% for element in value %} {{ (element.value)|round|int }}{% if not loop.last %},{% endif %} {% endfor %}]</td>
{% endif %}
{% endfor %}
{% for key,value in dictionary_mcqsr.items() %}
{% if key == j %}
<td>[{% for element in value %} {{ (element.value)|round|int }}{% if not loop.last %},{% endif %} {% endfor %}]</td>
{% endif %}
{% endfor %}
You can use sort and round filter in Jinja.
{% for key, value in dictionary_num.items() %}
{% if key == j%}
<td>
[
{% for element in value %}
{{ element.value | round | int }}
{% if not loop.last%}
,
{% endif %}
{% endfor %}
]
{% endif %}
{% endfor %}
{% for key, value in dictionary_mcqsr.items() %}
{% if key == j%}
<td>
[
{% for element in value | sort %}
{{ element.value | round | int }}
{% if not loop.last%}
,
{% endif %}
{% endfor %}
]
{% endif %}
{% endfor %}

Jinja add automatic indentation

In ansible I use a template written in jinja2,
I have an inner for loop which automatically adds space to my config file, which i do not want to.
stick store-response payload_lv(43,1) if serverhello
option ssl-hello-chk
{% set count = 1 %}
{% for ip in sv.ips %}
server server{{ count }} {{ ip }}:443 check
{% set count = count + 1 %}
{% endfor %}
Result is
stick store-response payload_lv(43,1) if serverhello
option ssl-hello-chk
server server1 10.2.0.16:443 check
server server2 10.2.0.20:443 check
Add this line at the top of your template, will preserve the indentation
#jinja2: trim_blocks: True, lstrip_blocks: True

jupyter nbconvert register jinja filter

For the jupyter sas_kernel I'm writing a template for nbconvert (version 4.2.0). The basic layout is as follows:
{%- extends 'null.tpl' -%}
{# Comment block #}
{% block header %}
/* coding: SAS utf-8 */
{% endblock header %}
{% block in_prompt %}
/* In[{{ cell.execution_count if cell.execution_count else ' ' }}]: */
{% endblock in_prompt %}
{% block input %}
{{ cell.source | ipython2python | sas_magic_regex }}
{% endblock input %}
{% block markdowncell scoped %}
/*
{{ cell.source | comment_lines(prefix='*') }}
*/
{% endblock markdowncell %}
In the block input section I need to use a regular expression to process the text from ipython2python mostly to deal with magic calls. I couldn't see a regex function from the string filers
so I wrote my own:
def sas_magic_regex(string:str) -> str:
""" Replace Python Magics with a commented out version or the SAS x command """
sas_magic=re.compile(r'^get_ipython.*?(\w+)\(\'(.*)\'\)')
newStr=sas_magic.search(string)
if len(newStr.groups())==2:
if newStr.group(1)=='system':
return 'x ' + newStr.group(2) + ';'
else:
return '/* Jupyter magic ' + string + ' */'
env = Environment()
env.filters['sas_magic_regex'] = sas_magic_regex
The problem is that nbconvert can't find this jinja filter.
I get this error message: jinja2.exceptions.TemplateAssertionError: no filter named 'sas_magic_regex'

Include different file in Jekyll depending on the locale

I'm trying to create my first Jekyll website and I'm encountering a problem designing the i18n part.
The different articles will be totally rewritten for each language, so each one will be a different post, no problems here. I actually have more difficulties with the text in my layout / includes.
Typically, for the menu, I was thinking doing something along these lines:
{% capture menu_location %}menu.{{ lang }}.html{% endcapture %}
{% include menu_location %}
like suggested here. But this gives me the following error :
Included file 'menu_location' not
found in _includes directory
Is it possible to use a variable for the include tag ? Do you have any other idea how I can do this ?
Thanks !
PS: Even if I have only 3 languages in mind for the moment, I won't do it with an if / elseif / else syntax ;)
What you are trying to do isn't possible without modifying Jekyll. The include filter they defined treats its parameter as a string, not as an expression.
I've created a couple bilingual sites with Jekyll in the past. I found that the cleanest solution was often storing locale-dependant variables inside _config.yml, and reference it when needed.
# _config.yml
...
locales:
en:
welcome_message: "Welcome to my site"
...
es:
welcome_message: "Bienvenido a mi sitio"
For every page I render, I need to know its locale. The simplest way of doing that is adding a locale field on the top yaml that page:
---
# a page or post (for example index.html)
...
locale: en
---
You can then get the current page locale by doing page.locale.
If you have divided your site in folders (/en/ for english, /es/ for spanish, and so on, you can use the page url to calculate the locale, so that you don't need to specify the locale on each page:
{% capture locale %}{{ page.url | truncate: 3, "" | remove: "/" }}{% endcapture %}
{% if locale == "" %}{% assign locale = "en" %}{% endif %}
For a page like /en/blog/, locale will be 'en'; for /fr/le-blog, it will be 'fr'. You will have to use that line in all the layouts and includes that need to use the locale, though.
You can then get localized texts like this:
{{ site.locales[locale].welcome_message }}
Or, if you prefer page.locale:
{{ site.locales[page.locale].welcome_message }}
Menus are the same; you can generate them from _config.yml, too:
# _config.yml
...
locales:
en:
nav:
- text: Welcome
url: /en/welcome
- text: Blog
url: /en/blog
layout: post
...
es:
nav:
- text: Bienvenido
url: /es/bienvenido
- text: Blog
url: /es/blog
layout: post
...
Then you can have a single menu.html that generates the right menu depending on the page locale:
{% capture locale %}{{ page.url | truncate: 3, "" | remove: "/" }}{% endcapture %}
{% if locale == "" %}{% assign locale = "en" %}{% endif %}
<nav>
<ul class="nav">
{% for link in site.locales[locale].nav %}
{% assign current = nil %}
{% if page.url == link.url or page.layout == link.layout %}
{% assign current = 'current' %}
{% endif %}
<li class="{% if forloop.first %}first{% endif %} {{ current }} {% if forloop.last %}last{% endif %}">
<a class="{{ current }}" href="{{ link.url }}">{{ link.text }}</a>
</li>
{% endfor %}
</ul>
</nav>
For using ´page.locale´ instead of ´locale´, just remove the first two lines and replace site.locales[locale].nav by site.locales[page.locale].nav
I hope this helps.
Regards!
How about creating a filter, wouldn't it be simpler? Something like:
<li>{{some_text_id|localize</li>

Resources