## Machine Learning for Language Detection in Python with scikit-learn

I recently played around with scraping public Telegram channels and finally want to do something with the data. In particular, I wanted to play around with Named Entity Recognition (NER) and Sentiment Analysis on the content of the Telegram posts. I think this will probably work much better when you do it for each language separately. So before we answer the question "what's this post about?" or "how angry is the person in the post?", let's answer the question "what language is the post in?" first. And we'll do it while warming up with that hole machine learning (ML) topic. Be aware: I have no idea what I'm doing. Do you want to know more?

## Prettier Struct Definitions for Python CTypes

I was writing some Python code that uses [ctypes][] for interfacing with the [Windows ToolHelp API](https://docs.microsoft.com/en-us/windows/win32/api/_toolhelp/). Specifically, I had to [define a Python equivalent][structs] for the [PROCESSENTRY32](https://docs.microsoft.com/en-us/windows/win32/api/tlhelp32/ns-tlhelp32-processentry32) struct. Now, the [ctypes][] [struct definitions][structs] are a little annoying because they do not give you type hints. You can add the type hints _on top_ of the _fields_ attribute but that looks a little silly because you _literally_ [write everything twice][wet]. In Python 3.7+1, you can solve this using a metaclass, and one stigma of metaclasses is that they have very few applications, so I decided to blog about this one:
class FieldsFromTypeHints(type(ctypes.Structure)):
def __new__(cls, name, bases, namespace):
from typing import get_type_hints
class AnnotationDummy:
__annotations__ = namespace.get('__annotations__', {})
annotations = get_type_hints(AnnotationDummy)
namespace['_fields_'] = list(annotations.items())
return type(ctypes.Structure).__new__(cls, name, bases, namespace)

and now you can write:
class PROCESSENTRY32(ctypes.Structure, metaclass=FieldsFromTypeHints):
dwSize              : ctypes.c_uint32
cntUsage            : ctypes.c_uint32
th32ProcessID       : ctypes.c_uint32
th32DefaultHeapID   : ctypes.POINTER(ctypes.c_ulong)
th32ModuleID        : ctypes.c_uint32
th32ParentProcessID : ctypes.c_uint32
pcPriClassBase      : ctypes.c_long
dwFlags             : ctypes.c_uint32
szExeFile           : ctypes.c_char * 260

The metaclass simply deduces the _fields_ attribute of the new class for you before the class is _even created_. It might make sense to think of metaclasses as "class creation hooks", i.e. you get to modify what you have written before the class is actually being defined. The following bit is just to be compatible with [PEP-563](https://www.python.org/dev/peps/pep-0563/):
        from typing import get_type_hints
class AnnotationDummy:
__annotations__ = namespace.get('__annotations__', {})
annotations = get_type_hints(AnnotationDummy)

And then, the following is the actual magic line:
        namespace['_fields_'] = list(annotations.items())

This adds the attribute _fields_ before the class is created. [ctypes]: https://docs.python.org/3/library/ctypes.html [structs]: https://docs.python.org/3/library/ctypes.html#structures-and-unions [wet]: https://en.wikipedia.org/wiki/Don%27t_repeat_yourself Do you want to know what I was coding?
1. Starting in Python 3.7, dictionaries keep the insertion order. []

## Java Test Coverage and The Missing Path

I started learning JUnit 4 and encountered the message "1 of 4 branches missed." when testing a code fragment like x && y where x and y depend on each other in a particular way. Here is my journey and a "solution". You want to know more? !

## Wrapping integers in Python with Metaclassing

I often require a wrapping integer type in Python, by which I actually mean a subclass of int where all operations are performed modulo some constant number $N$. There are two main use cases for this: 1. Working in a finite field for some cryptographic stuff, or solving problems on [Project Euler](https://projecteuler.net/about). 2. Having Python integers behave like machine registers (8, 16, 32, 64, even 128 bits - you name it.) I decided to solve this once and for all and wrote the integer wrapper class to end all integer wrapper classes. I also managed to keep it rather compact by using *»gasp«* metaclassing. Do you want to know more?

## Fast XOR in Python

As I [have hinted at before](/2017/09/20/just-some-friendly-advice/), the [PyCrypto library](https://www.dlitz.net/software/pycrypto/) [seems to be dead](https://github.com/dlitz/pycrypto/issues/173). The [PyCryptodome](https://www.pycryptodome.org/en/latest/) library is a fork that is promising because it is maintained and works in Python 3, but they have a bit of a finger-wagging attitude which sometimes means that you have to fight the library a bit:
>>> from Crypto.Cipher import ARC4
>>> cipher = ARC4.new(B'funk')
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "C:\Python37\lib\site-packages\Crypto\Cipher\ARC4.py", line 132, in new
return ARC4Cipher(key, *args, **kwargs)
File "C:\Python37\lib\site-packages\Crypto\Cipher\ARC4.py", line 57, in __init__
len(key))
ValueError: Incorrect ARC4 key length (4 bytes)
>>> ARC4.key_size = range(1,257)
>>> ARC4.new(B'funk').decrypt( ARC4.new(B'funk').encrypt( B'Hello World' ))
b'Hello World'

They certainly mean well, but the library is no place to impose security standards, in my opinion. In malware research for example, we often have to verbatim copy the appalling use of certain ciphers, like ARC4 with a 4-byte key. It happens all the time! I have been particularly struggling with [the removal of the XOR cipher](https://pycryptodome.readthedocs.io/en/latest/src/vs_pycrypto.html). The XOR implementation of PyCrypto was very fast, and in this article I will both benchmark how fast exactly it was and give you a drop-in replacement which degrades gracefully based on your options. Do you want to know more?

## Get the MSDN Library for offline use

So you develop in [Microsoft Visual Studio Community Edition](https://www.visualstudio.com/de/vs/community/) and you long for the old days when there was a way to get the [MSDN Library](https://msdn.microsoft.com/en-us/library/) as an offline help file? Fear not, you still can. Open Visual Studio, type Ctrl+Q to open the quick access bar, usually located in the upper right corner of your interface. Enter Help Viewer, it should yield one result by that name, marked as an *"individual component"*. Selecting that entry should allow you to download and install the Help Viewer. Now relaunch Visual Studio and start the Help Viewer via quick access in the same way. You will be prompted whether you want to download some *content* - and I bet you do.

## Python: post JSON data containing a datetime object with requests to flask

Spoiler: My main point in this post is not given away by the title. But first things first: What are all those words? Would you like to know more?

## Just some friendly advice

Part of me wants to write about all the [horror](https://www.bleepingcomputer.com/news/security/ten-malicious-libraries-found-on-pypi-python-package-index/) and [glory](https://crates.io) there is to be seen in package management, but quite frankly it'll take too long. Instead, I will just leave you with a tiny piece of advice. Here comes. If you are on Windows and you want to install a *legitimate* Python package (like for example [PyCryptodome](https://pypi.python.org/pypi/pycryptodome), because naturally you are **fully aware** that [PyCrypto is dead](https://github.com/dlitz/pycrypto/issues/173).), which in reality is a bottomless pit, at the center of which there is a C library, straight from hell - then maybe get the [Microsoft Compiler for Python](http://www.microsoft.com/en-us/download/details.aspx?id=44266) instead of, who knows, wasting hours or even days looking for a less reasonable solution. Credit, as so very often, [goes to stackoverflow](https://stackoverflow.com/a/27327236/1578458).

## GithubTodos

1. https://github.com/dear-github/dear-github []
2. https://github.com/isaacs/github/issues []
3. on a side note, tellmewhenitcloses.com is pretty handy to avoid too many notifications in the first place []
5. Using local storage is handy for people that are not very concerned about privacy and just use the cloud synchronization feature of their browser: the content of their ToDo list will then also just be synced to all their devices. []

## Denied Permissions: copying text

Obviously, my Bank does not provide a REST API to download the transactions happening on my accounts. After I asked for "machine parseable" data, they told me that I can download CSV files. Awesome! So I wrote a parser and lived happily ever after. Except that they change their CSV format without notice every few months and at some point they started to mix different encodings in the same file. So I lived unhapply, regularly fixing the script reading the CSV file. What did not change for around 10 years now are their banking statements. And this also holds for the PDFs you have to download, if you want to avoid getting them via snail mail (and paying for the postage of course). I decided to parse the PDFs instead and this went pretty well for may years... up to recently, when something changed (it may have been a software update of the parser I use or something on their side). Do you want to know more?

## Code Rust in Windows

I have started to [learn](http://rustbyexample.com/) [rust](http://rust-lang.org), and I am enjoying myself. This is a merge and update on the two fantastic blog posts on [how to setup Visual Studio Code for Rust](https://mobiarch.wordpress.com/2015/06/16/rust-using-visual-studio-code/) and [how to enable debugging](https://sherryummen.in/2016/09/02/debugging-rust-on-windows-using-visual-studio-code/). In my personal opinion, from among [all the available IDE solutions for rust](https://areweideyet.com/), this is the best. Do you want to know more?

## Readable Git Annex Unused

So unless you know and use [git annex](https://git-annex.branchable.com/), this is not going to be very useful for you. Check it out, though. It's pretty cool. Unless you are on Windows. In that case it's hell. Anyway, I wrote a script to help me figure out the output of [git annex unused](http://manpages.ubuntu.com/manpages/wily/man1/git-annex-unused.1.html). In short, it tells you what those files used to be called before you lost them. Script:
#!/usr/local/bin/python3.5
import re, sys, os
from subprocess import Popen,PIPE
FNULL = open(os.devnull, 'w')

def seeker(s):
process = Popen(["git", "log", "--stat", "-S", s], stderr=FNULL, stdout=PIPE)
match = re.search(r"(([ -~]*\/)*[ -~]*)\|", log)
if not match: return ''
else: return match.group(1).strip()

if len(sys.argv)>1:
print( seeker(sys.argv[1]) )
else:
clist = []
while True:
crawl = input().strip()
if not crawl: break
crawl = crawl.split()
clist.append(crawl)
for x in clist:
print('%s : %s' % ( x[0], seeker(x[1]) ) )

You can either call it with one argument which should be a key, or if you call it with no argument, it expects you to paste the list of results you got from git annex unused into stdin. It then goes through the list and tells you the corresponding filename for each key. This script is phenomenally stupid in that it does quite a terrible regular expression search on the output of git log and returns the first match it finds. Sue me, it works pretty well at my end.

## Binary Determinantal Complexity

My good friend and colleague Christian Ikenmeyer and I wrote this cute preprint about polynomials and how they can be written as the determinant of a matrix with entries equal to zero, one and indeterminantes. Go ahead and read it if you know even just a little math, it's quite straightforward. The algorithm described in section 3 has been implemented and you can download the code from my website at the TU Berlin. Compilation instructions are in ptest.c, but you will need to get nauty to perform the entire computerized proof.