RSS blog page - 1/6 ==> Blog

blog page - 1/6

Nbconvert requires recent version of latex


The latest version of nbconvert which converts notebooks into latex requires recent version latex > 2016. To install a more recent version on debian:




jupyter_sphinx was recently updated and now includes a directive which lets the user run a piece of code and even modify it before running.


Slow imports and code to investigate


I notice that the following code was very slow:


nbconvert, tornado, incompatibilities


tornado was recently released in version 6.0. It introduced an incompabilities with all existing versions of nbconvert (<= 0.4.1).


Missing signature for functions


The signature of a function is not always available in Python. The buildin functions do not follow the same pattern as functions written in Python but they provide a backup plan with the attribute __text_signature__:


Upload a single binary on PyPI


The packages for multiple distribution are built on different machine but the upload takes place on a single machine. I retrieved all available builds and moved them to the dist folder. The file .pypirc was saved into my home with the following content:


Check RST syntax


It is usually a pain to discover I made an error in a formula while I’m writing documentation. It fails quite long after after the unit tests started. The documentation is generated after the unit test pass. I also use a lot .. runpython:: (see RunPythonDirective) to run pieces of code inside the documentation. It is quite annoying to discover it fails long after. So I creates a unit test which can be used to compile a single page of the documentation : test_doc_page. I modify the page name when I have some doubt or I move to another one.


Long names on Windows


One of my unit test was failing due to a weird error in sphinx-gallery. The module could not write a file on Windows because its name was longer than 260 which still seems to be a limit of the system.


pip 9.0.2 and issue with pip._vendor.urllib3.contrib


pip was recently updated and the new version produces a weird error:


Replace pdflatex by xelatex


pdflatex does not natively support utf-8 (no fixed size characters) but xelatex does (see Swapping from pdfLaTeX to XeLaTex). I tried a couple of past in the past with no success and I finally remove all accents from pdf coming from notebooks. It seems to work now.


RSS blog page - 1/6 ==> 2018-05 (1) 2018-08 (1) 2019-03 (2) 2019-07 (1) 2019-11 (1)