Creating reports with litebird_sim

The LiteBIRD simulation framework has the ability to automatically produce reports in the output directory specified during the creation of a Simulation object.

The framework uses the Jinja2 templating library to render the text, and the Python-Markdown library to convert the text to HTML.

The usage of the facility is easy:

import litebird_sim as lbs
import matplotlib.pylab as plt

# Create some plot
data_points = [0, 1, 2, 3]
plt.plot(data_points)
fig = plt.gcf()

sim = lbs.Simulation(base_dir="./report_example", random_seed=12345)
sim.append_to_text("""

## Section name

Here is some Markdown text written in *italic* and in **bold**.

Here is a list:

- One
- Two
- Three

The value of the variable `foo` is {{ foo }}.

And here is a figure:

![](myfigure.png)
""",
    figures=[(fig, "myfigure.png")],
    foo=123,
)

The method Simulation.append_to_report() accepts arbitrary keywords, which are substituted in the string whenever the form {{ … }} is found. Thus, the string The value of the variable `foo` is {{ foo }} is converted into The value of the variable `foo` is 123. The keyword figures is however interpreted directly by Simulation.append_to_report(): it must be a list of 2-tuples of the form (figure, filename), and it triggers the save of the Matplotlib figure into a file with the specified name; you can refer to the file using Markdown’s usual syntax ![](filename), as shown in the example above.

Including tables

Reports can include tables as well. They are formatted using the ASCII characters | and -, as in the following example:

sim.append_to_report("""

Here is a table:

Column A   | Column B
---------- | ------------
Item 1 | Value 1
Item 2 | Value 2
Item 3 | Value 3
""")

If you are going to write the rows of the table using Jinja2’s for loops, be sure to end the loop with {% endfor -%} instead of {% endfor %}, in order not to leave empty lines between consecutive rows.

Converting the report to other formats

It might be the case that the report produced by your script contains some complex table/paragraphs that you would like to include in an article written in LaTeX or Microsoft Word. You can easily convert the report into one of these formats using Pandoc.

The Simulation Framework saves the Markdown source text of the report in the same directory as report.html, and you can ask Markdown to convert to some other format than HTML. The following shell commands will produce files report.tex (LaTeX source), report.docx (Microsoft Word file), and report.pdf (PDF file produced using LaTeX):

$ pandoc -f markdown -t latex -o report.tex report.md
$ pandoc -f markdown -t docx -o report.docx report.md
$ pandoc -f markdown -t latex -o report.pdf report.md

Any image included in the report is saved in the same directory as report.md and report.html, so you can take and reuse them. To ease the possibility to include images in the documents, you should stick to the PNG image format, as it is easily embeddable in HTML, pdfLaTeX/xeLaTeX/luaLaTeX, and Microsoft Word files. The alternative is to manually convert the image files to other formats.

If the report contains mathematical equations, things get a little trickier. You must create a text file with your favourite editor and copy the following code in it (the file is available in the Python distribution as litebird_sim/misc/pandoc-filter.lua):

-- mode: lua

-- Pass the following option to pandoc:
--
--   pandoc --lua-filter gitlab-math.lua ...
--
-- to use this filter

function Math(el)
   if el.mathtype == "InlineMath" then
      if el.text:sub(1,1) == '`' and el.text:sub(#el.text) == '`' then
         local text = el.text:sub(2,#el.text-1)
         return pandoc.Math(el.mathtype, text)
      else
         local cont = pandoc.read(el.text)
         return { pandoc.Str("$") } .. cont.blocks[1].content .. { pandoc.Str("$") }
      end
   end
end

function CodeBlock(el)
   if el.classes[1] == "math" then
      return pandoc.Para({ pandoc.Math("DisplayMath", el.text) })
   end
end

and save it somewhere in your computer; let’s suppose that the path is /my/scripts/litebird-sim-pandoc.lua (better to use the extension .lua, as this is a Lua script). Then, run pandoc with the option --lua-filter /my/scripts/litebird-sim-pandoc.lua and the equations will be converted properly. The following image shows the result of running the filter on a report generated by an analysis script to convert the report in a docx file:

And here is a screenshot of one page of the PDF file that has been produced by Pandoc by converting the same document to a LaTeX file:

Converting reports to LaTeX is particularly handy when you want to re-use tables written in the report, as these are notoriously a pain to write manually in LaTeX.

A few stylistical tips

Template files

If the report is getting larger and larger, it is advisable to move the string passed to Simulation.append_to_report() in a separate file and load it at runtime. (This is what we do in the source code of the framework, see the folder litebird_sim/templates.) For instance, you could put the following text in a file my_template_report.md:

Here is some number: {{ "0.2f" | format(val) }}
Blah blah blah

Here follows a few very long paragraphs...

And here some more text...

and then you would load it in your script in the following way:

with open("my_template_report.md", "rt") as inpf:
    template = "".join(inpf.readlines())

sim.append_to_report(template, val=value)

In this way, you can remove much clutter from the Python file, and it is easier for people who want to improve the report to contribute, as they do not need to understand Python to do it.

Where to put the formatting logic

You should always try to put the formatting logic of the report within the string passed to Simulation.append_to_report(), possibly using Jinja2’s filters, instead of formatting the arguments in your Python script. Thus, instead of doing this:

sim.append_to_report(
    "Here is some number: {{ val }}",
    val="%0.2f" % value,    # Don't do this!
)

you should do this:

# Ok, put the formatting logic into the string and rely
# on Jinja2's filter named "format"
sim.append_to_report(
    'Here is some number: {{ "0.2f" | format(val) }}',
    val=value,
)

The reason is that if you keep the template in a separate Markdown file, as explained above, it is better to put whatever concerns the representation of the values in the template instead of the code. For instance, at some point you might revise the layout of your report by adding space, and this might enable you to write some date field as February, 8th 2022 instead of 22-02-08; this kind of change should affect only the Markdown template and not the Python code, because it’s the template that was modified to enable more space.

Try to follow this tip even for short snippets: in the authors’ experience, it is often the case that short reports grow more and more as time passes, until at some point they are moved to separate Markdown files!

You can refer to Jinja2’s documentation for a complete list of filters, but here is a selection of the most useful ones:

• {{ str | format(value) }} formats value like the C function printf(str, value)

• {{ var | round(n) }} rounds val to n digits

• {{ var | filesizeformat }} interprets var as a size in bytes and formats it in a human-readable format. Examples: 13 kB, 4.1 MB, etc.

• {{ var | upper }} converts var (a string) to uppercase

• {{ var | lower }} converts var (a string) to lowercase

• {{ var | capitalize }} capitalizes var (a string), so that jupiter is turned into Jupiter.