Pydfy: Turn DataFrames into PDF Reports

Pydfy makes creating beautiful PDF reports in Python as simple as you'd hope it to be. In practice, this is often not the case as:

• Not the whole team knows LaTeX
• The support for this in BI tooling is lacking
• Webpages are usually not tuned for PDF exports
• PDF builders are too complex to get something out the door by the end of the week

Installation

First, make sure tailwindcss is in your PATH, which for Linux means:

curl -sLO https://github.com/tailwindlabs/tailwindcss/releases/latest/download/tailwindcss-linux-x64
chmod +x tailwindcss-linux-x64
sudo mv tailwindcss-linux-x64 /usr/local/bin/tailwindcss

Check https://github.com/tailwindlabs/tailwindcss/releases/latest to find the matching binary for your system.

Also make sure to install Chromium:

apt-get update && apt-get install -y chromium-driver  # Example using apt-get
brew install --cask chromium                          # Homebrew example for macOS

Now run:

pip install pydfy

And you should be good to go. Optionally install the extra dependencies as well:

pip install "pydfy[pandas,matplotlib]"

Usage

Example

import pydfy.models as pf
import pandas as pd

sales_df = pd.read_csv("sales.csv")
orders_last_week = sales_df[sales_df["week"] == 10]
avg_price = sales_df["price"].mean()

pdf = pf.PDF(
    pf.Title(text="Sales Report"),
    [pf.Number(orders_last_week, title="Orders last week"), pf.Number(avg_price, title="Average price")],
    [pf.Table(sales_df, "Sales table"), sales_df.plot("hist")],
)
pdf.render()

The main entrypoint is defined, in this context, by the pf.PDF function, which accepts the rows of the PDF as arguments and wraps a pf._PDF model for ease of use. A row here can either be a single component or an iterable of components, as shown in the example above.

Overview of Components

Component(template_path=..., col_span=...)  # Abstract component, args apply to all other components

Table(rows=..., title=..., headers=...)
Number(number=..., tiele=...)

Image(path=...)
Figure(figure=...)

Title(text=...)
Section(text=...)
Paragraph(content=...)
PageBreak(...)

Output location

By default, this renders a PDF file called out.pdf in the current working directory. To control the location of the build files, you can do:

# Change the name of the rendered report and store the intermediate html and css files in the /build directory
pdf.render(out="report.pdf", build_dir="/build")

# Save everything in the /data directory
pdf.render(out="/data/report.pdf", build_dir="/data")

# Only save the pdf in the /data directory
os.environ["PYDFY_OUT"] = "/data/report.pdf"
pdf.render()

Custom Components

You can create custom components with custom html templates as follows:

from dataclasses import dataclass, field

import pydfy.models as pf


@dataclass
class TwoNumbers(pf.Component):
    number_1: int
    number_2: int
    title: str

    template_path: str = field(default="two_numbers.html", kw_only=True)

Where the contents of two_numbers.html is:

<div class="pf-two-numbers border-solid border border-gray-100">
    <div class="pf-two-numbers-title bg-gray-200 text-lg font-bold text-center align-top">{{ component.title }}</div>
    <div class="pf-two-numbers-content mt-2 text-center">
        <div class="grid grid-cols-2">
            <div class="text-3xl mb-2">{{ "{:,}".format(component.number_1) }}</div>
            <div class="text-3xl mb-2">{{ "{:,}".format(component.number_2) }}</div>
        </div>
    </div>
</div>

To see a working example, check out examples/custom.

Custom CSS

You can add custom CSS to the compilation step:

pdf.render(css="blue.css")

Content of my.css, where we need to copy the contents of base.css first:

/* Tailwind setup */
@tailwind base;  @tailwind components; @tailwind utilities;
@media print { body { -webkit-print-color-adjust: exact; } }
@media print {  html, body { width: 210mm; height: 297mm; font-size: 8pt; } }
@media print { thead {display: table-row-group;} }
@page { size: A4; margin: 1cm; }

.pf-number-content {
    color: blue;
}

To see a working example, check out examples/iris.

Using Docker

You could use Docker in the development process as follows:

docker build -t pydfy .
cd examples/iris
docker run -v $PWD:/pydfy pydfy > out.pdf  # The docker image is configured to run main.py and write to stdout

Contributing

poetry install --with dev --all-extras