PDF generation in python using wkhtmltopdf.
Wkhtmltopdf binaries are precompiled and included in the package making pydf easier to use, in particular this means pydf works on heroku.
Currently using wkhtmltopdf 0.12.6.1 r3 for Ubuntu 22.04 (jammy), requires Python 3.6+.
If you're not on Linux amd64: pydf comes bundled with a wkhtmltopdf binary which will only work on Linux amd64
architectures. If you're on another OS or architecture your mileage may vary, it is likely that you'll need to supply
your own wkhtmltopdf binary and point pydf towards it by setting the WKHTMLTOPDF_PATH
environment variable.
pip install python-pdf
import pydf
pdf = pydf.generate_pdf('<h1>this is html</h1>')
with open('test_doc.pdf', 'wb') as f:
f.write(pdf)
Generation of lots of documents with wkhtmltopdf can be slow as wkhtmltopdf can only generate one document
per process. To get round this pydf uses python 3's asyncio create_subprocess_exec
to generate multiple pdfs
at the same time. Thus the time taken to spin up processes doesn't slow you down.
from pathlib import Path
from pydf import AsyncPydf
async def generate_async():
apydf = AsyncPydf()
async def gen(i):
pdf_content = await apydf.generate_pdf('<h1>this is html</h1>')
Path(f'output_{i:03}.pdf').write_bytes(pdf_content)
coros = [gen(i) for i in range(50)]
await asyncio.gather(*coros)
loop = asyncio.get_event_loop()
loop.run_until_complete(generate_async())
See benchmarks/run.py for a full example.
Locally generating an entire invoice goes from 0.372s/pdf to 0.035s/pdf with the async model.
pydf is available as a docker image with a very simple http API for generating pdfs.
Simple POST
(or GET
with data if possible) you HTML data to /generate.pdf
.
Arguments can be passed using http headers; any header starting pdf-
or pdf_
will
have that prefix removed, be converted to lower case and passed to wkhtmltopdf.
For example:
docker run -rm -p 8000:80 -d samuelcolvin/pydf
curl -d '<h1>this is html</h1>' -H "pdf-orientation: landscape" http://localhost:8000/generate.pdf > created.pdf
open "created.pdf"
In docker compose:
services:
pdf:
image: samuelcolvin/pydf
Other services can then generate PDFs by making requests to pdf/generate.pdf
. Pretty cool.
generate_pdf(source, [**kwargs])
Generate a pdf from either a url or a html string.
After the html and url arguments all other arguments are passed straight to wkhtmltopdf
For details on extra arguments see the output of get_help() and get_extended_help()
All arguments whether specified or caught with extra_kwargs are
converted to command line args with '--' + original_name.replace('_', '-')
.
Arguments which are True are passed with no value eg. just --quiet, False and None arguments are missed, everything else is passed with str(value).
Arguments:
source
: html string to generate pdf from or url to getquiet
: boolgrayscale
: boollowquality
: boolmargin_bottom
: string eg. 10mmmargin_left
: string eg. 10mmmargin_right
: string eg. 10mmmargin_top
: string eg. 10mmorientation
: Portrait or Landscapepage_height
: string eg. 10mmpage_width
: string eg. 10mmpage_size
: string: A4, Letter, etc.image_dpi
: int default 600image_quality
: int default 94extra_kwargs
: any exotic extra options for wkhtmltopdf
Returns string representing pdf
get_version()
Get version of pydf and wkhtmltopdf binary
get_help()
get help string from wkhtmltopdf binary uses -h command line option
get_extended_help()
get extended help string from wkhtmltopdf binary uses -H command line option
execute_wk(*args)
Low level function to call wkhtmltopdf, arguments are added to wkhtmltopdf binary and passed to subprocess with not processing.
If you are deploying onto Heroku, then you will need to install a couple of dependencies before WKHTMLTOPDF will work.
Add the Heroku buildpack https://buildpack-registry.s3.amazonaws.com/buildpacks/heroku-community/apt.tgz
Then create an Aptfile in your root directory with the dependencies: