Planeta PythonBrasil

I write a lot of Telegram bots using the library python-telegram-bot. Writing Telegram bots is fun, but you will also need someplace to host them.

I personally like the new Google Cloud Run; or run, for short, is perfect because it has a “gorgeous” free quota that should be mostly sufficient to host your bots, also, and is it super simple to deploy and get running.

To create Telegram bots, first, you need to talk to BotFather and get a TOKEN.

Secondly, you need some coding. As I mentioned before, you can use python-telegram-bot to do your bots. Here is the documentation.

Code

Here is the base code that you will need to run on Cloud Run.

main.py

import os
import http

from flask import Flask, request
from werkzeug.wrappers import Response

from telegram import Bot, Update
from telegram.ext import Dispatcher, Filters, MessageHandler, CallbackContext

app = Flask(__name__)


def echo(update: Update, context: CallbackContext) -> None:
    update.message.reply_text(update.message.text)

bot = Bot(token=os.environ["TOKEN"])

dispatcher = Dispatcher(bot=bot, update_queue=None, workers=0)
dispatcher.add_handler(MessageHandler(Filters.text & ~Filters.command, echo))

@app.route("/", methods=["POST"])
def index() -> Response:
    dispatcher.process_update(
        Update.de_json(request.get_json(force=True), bot))

    return "", http.HTTPStatus.NO_CONTENT

requirements.txt

flask==1.1.2
gunicorn==20.0.4
python-telegram-bot==13.1

Dockerfile

FROM python:3.8-slim
ENV PYTHONUNBUFFERED True
WORKDIR /app
COPY *.txt .
RUN pip install --no-cache-dir --upgrade pip -r requirements.txt
COPY . ./

CMD exec gunicorn --bind :$PORT --workers 1 --threads 8 --timeout 0 main:app

Deployment

Finally, you need to deploy. You can do it in a single step, but first, let’s run the command below to set the default region (optionally).

gcloud config set run/region us-central1

Then deploy to Cloud Run:

gcloud beta run deploy your-bot-name \
    --source . \
    --set-env-vars TOKEN=your-telegram-bot-token \
    --platform managed \
    --allow-unauthenticated \
    --project your-project-name

After this, you will receive a public URL of your run, and you will need to set the Telegram bot webHook using cURL

curl "https://api.telegram.org/botYOUR-BOT:TOKEN/setWebhook?url=https://your-bot-name-uuid-uc.a.run.app"

You should replace the YOUR-BOT:TOKEN by the bot’s token and the public URL of your Cloud Run.

This should be enough.

Why?

Google Cloud Storage is cheaper, and you pay only for what you use than Google One. Also, you can erase any photo, and you still have a copy of that.

Installation

Create a Compute Engine (a VM).

If you choose Ubuntu, first of all, remove snap

sudo apt autoremove --purge snapd
sudo rm -rf /var/cache/snapd/
rm -rf ~/snap

Install gcsfuse or follow the official instructions.

export GCSFUSE_REPO=gcsfuse-`lsb_release -c -s`
echo "deb http://packages.cloud.google.com/apt $GCSFUSE_REPO main" | sudo tee /etc/apt/sources.list.d/gcsfuse.list
curl https://packages.cloud.google.com/apt/doc/apt-key.gpg | sudo apt-key add -

sudo apt-get update
sudo apt-get install gcsfuse

On Google Cloud console create a bucket of the type Nearline, in my case the name of the bucket is tank1, then back to your VM and create a dir with the same name of the bucket.

mkdir name-of-your-bucket

Now install gphotos-sync.

sudo apt install -y python3-pip
pip3 install gphotos-sync

I created a small Python script to deal with multiple Google accounts. I’ll explain later how it works.

cat <<EOF > /home/ubuntu/synchronize.py
#!/usr/bin/env python3

import os
import sys
import subprocess
from pathlib import Path

import requests


home = Path(os.path.expanduser("~")) / "tank1/photos"

args = [
  "--ntfs",
  "--retry-download",
  "--skip-albums",
  "--photos-path", ".",
  "--log-level", "DEBUG",
]

env = os.environ.copy()
env["LC_ALL"] = "en_US.UTF-8"

for p in home.glob("*/*"):
  subprocess.run(["/home/ubuntu/.local/bin/gphotos-sync", *args, str(p.relative_to(home))], check=True, cwd=home, env=env, stdout=sys.stdout, stderr=subprocess.STDOUT)

# I use healthchecks.io to alert me if the script has stopped work
url = "https://hc-ping.com/uuid4"
response = requests.get(url, timeout=60)
response.raise_for_status()
EOF

Give execute permission.

chmod u+x synchronize.py

Now let’s create some systemd scripts.

sudo su

Let’s create a service to gcsfuse, responsible to mount the bucket locally using the FUSE.

cat <<EOF >/etc/systemd/system/gcsfuse.service 
# Script stolen from https://gist.github.com/craigafinch/292f98618f8eadc33e9633e6e3b54c05
[Unit]
Description=Google Cloud Storage FUSE mounter
After=local-fs.target network-online.target google.service sys-fs-fuse-connections.mount
Before=shutdown.target

[Service]
Type=forking
User=ubuntu
ExecStart=/bin/gcsfuse tank1 /home/ubuntu/tank1
ExecStop=/bin/fusermount -u /home/ubuntu/tank1
Restart=always

[Install]
WantedBy=multi-user.target
EOF

Enable and start the service:

systemctl enable gcsfuse.service
systemctl start gcsfuse.service

cat <<EOF >/etc/systemd/system/gphotos-sync.service 
[Unit]
Description=Run gphotos-sync for each account

[Service]
User=ubuntu
ExecStart=/home/ubuntu/synchronize.py
EOF

And enable the service.

systemctl enable gphotos-sync.service

Now let’s create a timer to run 1 minute after the boot the gphotos-sync.service with gcsfuse.service as dependency.

cat <<EOF >/etc/systemd/system/gphotos-sync.timer 
[Unit]
Description=Run gphotos sync service weekly
Requires=gcsfuse.service

[Timer]
OnBootSec=1min
Unit=gphotos-sync.service

[Install]
WantedBy=timers.target
EOF

systemctl enable gphotos-sync.timer
systemctl start gphotos-sync.timer

exit (back to ubuntu user)

Now follow https://docs.google.com/document/d/1ck1679H8ifmZ_4eVbDeD_-jezIcZ-j6MlaNaeQiz7y0/edit to get a client_secret.json to use with gphotos-sync.

mkdir -p /home/ubuntu/.config/gphotos-sync/
# Copy the contents of the json to the file bellow
vim /home/ubuntu/.config/gphotos-sync/client_secret.json 

Testing

Due to an issue with gcsfuse, I was unable to create the backup dir directly on the bucket. The workaround is to create a temp directory and start the gphotos-sync manually first.

mkdir -p ~/temp/username/0
cd ~/temp
gphotos-sync --ntfs --skip-albums --photos-path . username/0
# gphotos-sync will ask for a token, paste it and CTRL-C to stop the download of photos.
cp ~/temp/username/ ~/tank1/photos/username

Verify if it is working.

./synchronize.py

After executing the command above, the script should start the backup. You can wait until it finishes or continue to the steps below.

Schedule startup and shutdown of the VM

The content below is based on and simplified version of Scheduling compute instances with Cloud Scheduler by Google

Back to your VM and add the label runtime with the value weekly, this is needed by the function below to know which instances should be started or shutdown.

Create a new directory, in my case, I will call functions and add two files:

index.js

const Compute = require('@google-cloud/compute');
const compute = new Compute();

exports.startInstancePubSub = async (event, context, callback) => {
  try {
    const payload = JSON.parse(Buffer.from(event.data, 'base64').toString());
    const options = {filter: `labels.${payload.label}`};
    const [vms] = await compute.getVMs(options);
    await Promise.all(
      vms.map(async instance => {
        if (payload.zone === instance.zone.id) {
          const [operation] = await compute
            .zone(payload.zone)
            .vm(instance.name)
            .start();

          return operation.promise();
        }
      })
    );

    const message = 'Successfully started instance(s)';
    console.log(message);
    callback(null, message);
  } catch (err) {
    console.log(err);
    callback(err);
  }
};

exports.stopInstancePubSub = async (event, context, callback) => {
  try {
    const payload = JSON.parse(Buffer.from(event.data, 'base64').toString());
    const options = {filter: `labels.${payload.label}`};
    const [vms] = await compute.getVMs(options);
    await Promise.all(
      vms.map(async instance => {
        if (payload.zone === instance.zone.id) {
          const [operation] = await compute
            .zone(payload.zone)
            .vm(instance.name)
            .stop();

          return operation.promise();
        } else {
          return Promise.resolve();
        }
      })
    );

    const message = 'Successfully stopped instance(s)';
    console.log(message);
    callback(null, message);
  } catch (err) {
    console.log(err);
    callback(err);
  }
};

And

package.json

{
  "main": "index.js",
  "private": true,
  "dependencies": {
    "@google-cloud/compute": "^2.4.1"
  }
}

Create a PubSub topic to start the instance.

gcloud pubsub topics create start-instance-event

Now deploy the startInstancePubSub function

gcloud functions deploy startInstancePubSub \
    --trigger-topic start-instance-event \
    --runtime nodejs12 \
    --allow-unauthenticated

And another PubSub topic to stop the instance.

gcloud pubsub topics create stop-instance-event

And the stopInstancePubSub function

gcloud functions deploy stopInstancePubSub \
    --trigger-topic stop-instance-event \
    --runtime nodejs12 \
    --allow-unauthenticated

And finally, let’s create two Cloud Scheduler to publish on the topics on Sunday and Monday at midnight.

gcloud beta scheduler jobs create pubsub startup-weekly-instances \
    --schedule '0 0 * * SUN' \
    --topic start-instance-event \
    --message-body '{"zone":"us-central1-a", "label":"runtime=weekly"}' \
    --time-zone 'America/Sao_Paulo'

gcloud beta scheduler jobs create pubsub shutdown-weekly-instances \
    --schedule '0 0 * * MON' \
    --topic stop-instance-event \
    --message-body '{"zone":"us-central1-a", "label":"runtime=weekly"}' \
    --time-zone 'America/Sao_Paulo'

After this setup, your VM will start every Sunday, backup all your photos of all accounts and shutdown on Monday.

I was using Scrapy to crawl some websites and mirror their content into a new one and at the same time, generate beautiful and unique URLs based on the title, but the title can appear repeated! So I added part of the original URL in base36 as uniqueness guarantees.

In the URL I wanted the title without special symbols, only ASCII and at the end a unique and short inditifier, and part of the result of the SHA-256 of the URL in base36.

class PreparePipeline():
  def process_item(self, item, spider):
    title = item.get("title")
    if title is None:
      raise DropItem(f"No title were found on item: {item}.")

    url = item["url"]

    N = 4
    sha256 = hashlib.sha256(url.encode()).digest()
    sliced = int.from_bytes(
        memoryview(sha256)[:N].tobytes(), byteorder=sys.byteorder)
    uid = base36.dumps(sliced)

    strip = str.strip
    lower = str.lower
    split = str.split
    deunicode = lambda n: normalize("NFD", n).encode("ascii", "ignore").decode("utf-8")
    trashout = lambda n: re.sub(r"[.,-@/\\|*]", " ", n)
    functions = [strip, deunicode, trashout, lower, split]
    fragments = [
        *functools.reduce(
        lambda x, f: f(x), functions, title),
        uid,
    ]

    item["uid"] = "-".join(fragments)

    return item

For example, with the URL https://en.wikipedia.org/wiki/Déjà_vu and title Déjà vu - Wikipedia will result in: deja-vu-wikipedia-1q9i86k. Which is perfect for my use case.

I am starting a new series of small snippets of code which I think that maybe useful or inspiring for others.

Let’s suppose you have a pandas’ dataframe with a column named URL which one do you want to download.

The code below takes the advantage of the multi-core processing using the ThreadPoolExecutor with requests.

import multiprocessing
import concurrent.futures

from requests import Session
from requests.adapters import HTTPAdapter
from requests.packages.urllib3.util.retry import Retry

session = Session()

retry = Retry(connect=8, backoff_factor=0.5)
adapter = HTTPAdapter(max_retries=retry)
session.mount("http://", adapter)
session.mount("https://", adapter)


def download(url):
    filename = "/".join(["subdir", url.split("/")[-1]])

    with session.get(url, stream=True) as r:
        if not r.ok:
            return

        with open(filename, "wb") as f:
            for chunk in r.iter_content(chunk_size=8192):
                f.write(chunk)


def run(df, processes=multiprocessing.cpu_count() * 2):
    with concurrent.futures.ThreadPoolExecutor(processes) as pool:
        list(pool.map(download, df["url"]))


if __name__ == '__main__':
    df = pd.read_csv("download.csv")
    run(df)

Apresentação

Em algum momento, durante o seu processo de desenvolvimento com Django, pode ser que surja a necessidade de criar e restaurar o banco de dados da aplicação. Pensando nisso, resolvi fazer um pequeno tutorial, básico, de como realizar essa operação.

Nesse tutorial, usaremos o django-dbbackup, um pacote desenvolvido especificamente para isso.

Configurando nosso ambiente

Primeiro, partindo do início, vamos criar uma pasta para o nosso projeto e, nela, isolar o nosso ambiente de desenvolvimento usando uma virtualenv:

mkdir projeto_db && cd projeto_db #criando a pasta do nosso projeto

virtualenv -p python3.8 env && source env/bin/activate #criando e ativando a nossa virtualenv

Depois disso e com o nosso ambiente já ativo, vamos realizar os seguintes procedimentos:

pip install -U pip #com isso, atualizamos a verão do pip instalado

Instalando as dependências

Agora, vamos instalar o Django e o pacote que usaremos para fazer nossos backups.

pip install Django==3.1.2 #instalando o Django

pip install django-dbbackup #instalando o django-dbbackup

Criando e configurando projeto

Depois de instaladas nossas dependências, vamos criar o nosso projeto e configurar o nosso pacote nas configurações do Django.

django-admin startproject django_db . #dentro da nossa pasta projeto_db, criamos um projeto Django com o nome de django_db.

Depois de criado nosso projeto, vamos criar e popular o nosso banco de dados.

python manage.py migrate #com isso, sincronizamos o estado do banco de dados com o conjunto atual de modelos e migrações.

Criado nosso banco de dados, vamos criar um superusuário para podemos o painel admin do nosso projeto.

python manage.py createsuperuser

Perfeito. Já temos tudo que precisamos para executar nosso projeto. Para execução dele, é só fazermos:

python manage.py runserver

Você terá uma imagem assim do seu projeto:

Configurando o django-dbbackup

Dentro do seu projeto, vamos acessar o arquivo settings.py, como expresso abaixo:

django_db/
├── settings.py

Dentro desse arquivos iremos, primeiro, adiconar o django-dbbackup às apps do projeto:

INSTALLED_APPS = (
    ...
    'dbbackup',  # adicionando django-dbbackup
)

Depois de adicionado às apps, vamos dizer para o Django o que vamos salvar no backup e, depois, indicar a pasta para onde será encaminhado esse arquivo. Essa inserção deve ou pode ser feita no final do arquivo settings.py:

DBBACKUP_STORAGE = 'django.core.files.storage.FileSystemStorage' #o que salvar
DBBACKUP_STORAGE_OPTIONS = {'location': 'backups/'} # onde salvar

Percebam que dissemos para o Django salvar o backup na pasta backups, mas essa pasta ainda não existe no nosso projeto. Por isso, precisamos criá-la [fora da pasta do projeto]:

mkdir backups

Criando e restaurando nosso backup

Já temos tudo pronto. Agora, vamos criar o nosso primeiro backup:

python manage.py dbbackup

Depois de exetudado, será criado um arquivo -- no nosso exemplo, esse arquivo terá uma extensão .dump --, salvo na pasta backups. Esse arquivo contem todo backup do nosso banco de dados.

Para recuperarmos nosso banco, vamos supor que migramos nosso sistema de um servidor antigo para um novo e, por algum motivo, nossa base de dados foi corrompida, inviabilizando seu uso. Ou seja, estamos com o sistema/projeto sem banco de dados -- ou seja, exlua ou mova a a sua base dados .sqlite3 para que esse exemplo seja útil --, mas temos os backups. Com isso, vamos restaurar o banco:

python manage.py dbrestore

Prontinho, restauramos nosso banco de dados. O interessante do django-dbbackup, dentre outras coisas, é que ele gera os backups com datas e horários específicos, facilitando o processo de recuperação das informações mais recentes.

Por hoje é isso, pessoal. Até a próxima. ;)

(Ok a piada com seqtembro funciona melhor na versão em inglês, seqtember, mas simbora)

Por uma grande coincidência, obra do destino, ou nada disso, teremos um Setembro de 2020 repleto de eventos virtuais e gratuitos de alta qualidade sobre Qt e KDE.

Começando de 4 à 11 do referido mês teremos o Akademy 2020, o grande encontro mundial da comunidade KDE que esse ano, por motivos que todos sabemos, acontecerá de forma virtual. A programação do Akademy traz palestras, treinamentos, hacking sessions, discussões com foco em aplicações KDE específicas, e mais, reunindo hackers, designers, gerentes de projetos, tradutores, e colabores dos mais diversos segmentos para discutir e planejar o KDE e seus futuros passos.

E como falamos em KDE, por extensão, também falamos em Qt – afinal, grande parte das aplicações é escrita nesse framework. Portanto, mesmo que você trabalhe com Qt mas não use nada do KDE, vale a pena participar do evento – e também se perguntar “porque diabos não estou usando e desenvolvendo aplicações do KDE?”.

Um incentivo extra é que durante o Akademy, entre 7 e 11, acontecerá o Qt Desktop Days, evento da KDAB voltado para Qt no desktop (surpresa?). A programação preliminar já está disponível e será muito interessante ver os avanços da tecnologia em um campo que pode parecer menos sexy hoje em dia, por conta da muita atenção dada a projetos mobile ou embarcados, mas que pelo contrário, continua vibrante e recebendo muito investimento.

Após uma rápida pausa para um respiro, temos a Maratona Qt. Nosso amigo Sandro Andrade, professor do IFBA e colaborador de longa data do KDE, resolveu dedicar uma semana inteira, de 14 à 18 de setembro, para apresentar 5 tópicos sobre o Qt tratando de seus fundamentos e passos iniciais de cada um. O programa cobre QML, C++ e Qt, Qt no Android, no iOS, na web (sim!), computação gráfica e mesmo jogos! Extremamente recomendada pra todo mundo que conhece ou quer conhecer o framework.

A Maratona Qt vai servir como um esquenta para a QtCon Brasil 2020, esse ano também virtual. Em 26 e 27 de setembro o pessoal da qmob.solutions reunirá desenvolvedores Qt de vários países para apresentarem, entre outras coisas, trabalhos com Wayland, visão computacional e IA, análise de dados, Python, containers, prototipagem, embarcados, e outros, tudo envolvendo Qt! E também haverá uma apresentação sobre a próxima versão major da ferramenta, Qt 6.

Portanto pessoal, reservem este mês para uma grande imersão nos vários aspectos e possibilidades disponibilizadas pelo Qt.

Last time I described a way to search MAGs in metagenomes, and teased about interesting results. Let's dig in some of them!

I prepared a repo with the data and a notebook with the analysis I did in this post. You can also follow along in Binder, as well as do your own analysis!

Preparing some metadata

The supplemental materials for Tully et al include more details about each MAG, so let's download them. I prepared a small snakemake workflow to do that, as well as downloading information about the SRA datasets from Tara Oceans (the dataset used to generate the MAGs), as well as from Parks et al, which also generated MAGs from Tara Oceans. Feel free to include them in your analysis, but I was curious to find matches in other metagenomes.

Loading the data

The results from the MAG search are in a CSV file, with a column for the MAG name, another for the SRA dataset ID for the metagenome and a third column for the containment of the MAG in the metagenome. I also fixed the names to make it easier to query, and finally removed the Tara and Parks metagenomes (because we already knew they contained these MAGs).

This left us with 23,644 SRA metagenomes with matches, covering 2,291 of the 2,631 MAGs. These are results for a fairly low containment (10%), so if we limit to MAGs with more than 50% containment we still have 1,407 MAGs and 2,938 metagenomes left.

TOBG_NP-110, I choose you!

That's still a lot, so I decided to pick a candidate to check before doing any large scale analysis. I chose TOBG_NP-110 because there were many matches above 50% containment, and even some at 99%. Turns out it is also an Archaeal MAG that failed to be classified further than Phylum level (Euryarchaeota), with a 70.3% complete score in the original analysis. Oh, let me dissect the name a bit: TOBG is "Tara Ocean Binned Genome" and "NP" is North Pacific.

And so I went checking where the other metagenome matches came from. 5 of the 12 matches above 50% containment come from one study, SRP044185, with samples collected from a column of water in a station in Manzanillo, Mexico. Other 3 matches come from SRP003331, in the South Pacific ocean (in northern Chile). Another match, ERR3256923, also comes from the South Pacific.

What else can I do?

I'm curious to follow the refining MAGs tutorial from the Meren Lab and see where this goes, and especially in using spacegraphcats to extract neighborhoods from the MAG and better evaluate what is missing or if there are other interesting bits that the MAG generation methods ended up discarding.

So, for now that's it. But more important, I didn't want to sit on these results until there is a publication in press, especially when there are people that can do so much more with these, so I decided to make it all public. It is way more exciting to see this being used to know more about these organisms than me being the only one with access to this info.

And yesterday I saw this tweet by @DrJonathanRosa, saying:

I don’t know who told students that the goal of research is to find some previously undiscovered research topic, claim individual ownership over it, & fiercely protect it from theft, but that almost sounds like, well, colonialism, capitalism, & policing

Amen.

I want to run this with my data!

Next time. But we will have a discussion about scientific infrastructure and sustainability first =]

Comments?

• Thread on Twitter

(or: Top-down and bottom-up approaches for working around sourmash limitations)

In the last month I updated wort, the system I developed for computing sourmash signature for public genomic databases, and started calculating signatures for the metagenomes in the Sequence Read Archive. This is a more challenging subset than the microbial datasets I was doing previously, since there are around 534k datasets from metagenomic sources in the SRA, totalling 447 TB of data. Another problem is the size of the datasets, ranging from a couple of MB to 170 GB. Turns out that the workers I have in wort are very good for small-ish datasets, but I still need to figure out how to pull large datasets faster from the SRA, because the large ones take forever to process...

The good news is that I managed to calculate signatures for almost 402k of them ¹, which already let us work on some pretty exciting problems =]

Looking for MAGs in the SRA

Metagenome-assembled genomes are essential for studying organisms that are hard to isolate and culture in lab, especially for environmental metagenomes. Tully et al published 2,631 draft MAGs from 234 samples collected during the Tara Oceans expedition, and I wanted to check if they can also be found in other metagenomes besides the Tara Oceans ones. The idea is to extract the reads from these other matches and evaluate how the MAG can be improved, or at least evaluate what is missing in them. I choose to use environmental samples under the assumption they are easier to deposit on the SRA and have public access, but there are many human gut microbiomes in the SRA and this MAG search would work just fine with those too.

Moreover, I want to search for containment, and not similarity. The distinction is subtle, but similarity takes into account both datasets sizes (well, the size of the union of all elements in both datasets), while containment only considers the size of the query. This is relevant because the similarity of a MAG and a metagenome is going to be very small (and is symmetrical), but the containment of the MAG in the metagenome might be large (and is asymmetrical, since the containment of the metagenome in the MAG is likely very small because the metagenome is so much larger than the MAG).

The computational challenge: indexing and searching

sourmash signatures are a small fraction of the original size of the datasets, but when you have hundreds of thousands of them the collection ends up being pretty large too. More precisely, 825 GB large. That is way bigger than any index I ever built for sourmash, and it would also have pretty distinct characteristics than what we usually do: we tend to index genomes and run search (to find similar genomes) or gather (to decompose metagenomes into their constituent genomes), but for this MAG search I want to find which metagenomes have my MAG query above a certain containment threshold. Sort of a sourmash search --containment, but over thousands of metagenome signatures. The main benefit of an SBT index in this context is to avoid checking all signatures because we can prune the search early, but currently SBT indices need to be totally loaded in memory during sourmash index. I will have to do this in the medium term, but I want a solution NOW! =]

sourmash 3.4.0 introduced --from-file in many commands, and since I can't build an index I decided to use it to load signatures for the metagenomes. But... sourmash search tries to load all signatures in memory, and while I might be able to find a cluster machine with hundreds of GBs of RAM available, that's not very practical.

So, what to do?

The top-down solution: a snakemake workflow

I don't want to modify sourmash now, so why not make a workflow and use snakemake to run one sourmash search --containment for each metagenome? That means 402k tasks, but at least I can use batches and SLURM job arrays to submit reasonably-sized jobs to our HPC queue. After running all batches I summarized results for each task, and it worked well for a proof of concept.

But... it was still pretty resource intensive: each task was running one query MAG against one metagenome, and so each task needed to do all the overhead of starting the Python interpreter and parsing the query signature, which is exactly the same for all tasks. Extending it to support multiple queries to the same metagenome would involve duplicating tasks, and 402k metagenomes times 2,631 MAGs is... a very large number of jobs.

I also wanted to avoid clogging the job queues, which is not very nice to the other researchers using the cluster. This limited how many batches I could run in parallel...

The bottom-up solution: Rust to the rescue!

Thinking a bit more about the problem, here is another solution: what if we load all the MAGs in memory (as they will be queried frequently and are not that large), and then for each metagenome signature load it, perform all MAG queries, and then unload the metagenome signature from memory? This way we can control memory consumption (it's going to be proportional to all the MAG sizes plus the size of the largest metagenome) and can also efficiently parallelize the code because each task/metagenome is independent and the MAG signatures can be shared freely (since they are read-only).

This could be done with the sourmash Python API plus multiprocessing or some other parallelization approach (maybe dask?), but turns out that everything we need comes from the Rust API. Why not enjoy a bit of the fearless concurrency that is one of the major Rust goals?

The whole code ended up being 176 lines long, including command line parsing using strucopt and parallelizing the search using rayon and a multiple-producer, single-consumer channel to write results to an output (either the terminal or a file). This version took 11 hours to run, using less than 5GB of RAM and 32 processors, to search 2k MAGs against 402k metagenomes. And, bonus! It can also be parallelized again if you have multiple machines, so it potentially takes a bit more than an hour to run if you can allocate 10 batch jobs, with each batch 1/10 of the metagenome signatures.

So, is bottom-up always the better choice?

I would like to answer "Yes!", but bioinformatics software tends to be organized as command line interfaces, not as libraries. Libraries also tend to have even less documentation than CLIs, and this particular case is not a fair comparison because... Well, I wrote most of the library, and the Rust API is not that well documented for general use.

But I'm pretty happy with how the sourmash CLI is viable both for the top-down approach (and whatever workflow software you want to use) as well as how the Rust core worked for the bottom-up approach. I think the most important is having the option to choose which way to go, especially because now I can use the bottom-up approach to make the sourmash CLI and Python API better. The top-down approach is also way more accessible in general, because you can pick your favorite workflow software and use all the tricks you're comfortable with.

But, what about the results?!?!?!

Next time. But I did find MAGs with over 90% containment in very different locations, which is pretty exciting!

I also need to find a better way of distributing all these signature, because storing 4 TB of data in S3 is somewhat cheap, but transferring data is very expensive. All signatures are also available on IPFS, but I need more people to host them and share. Get in contact if you're interested in helping =]

And while I'm asking for help, any tips on pulling data faster from the SRA are greatly appreciated!

Comments?

• Thread on Twitter

Footnotes

1. pulling about a 100 TB in 3 days, which was pretty fun to see because I ended up DDoS myself because I couldn't download the generated sigs fast enough from the S3 bucket where they are temporarily stored =P ↩

sourmash 3.3 was released last week, and it is the first version supporting zipped databases. Here is my personal account of how that came to be =]

What is a sourmash database?

A sourmash database contains signatures (typically Scaled MinHash sketches built from genomic datasets) and an index for allowing efficient similarity and containment queries over these signatures. The two types of index are SBT, a hierarchical index that uses less memory by keeping data on disk, and LCA, an inverted index that uses more memory but is potentially faster. Indices are described as JSON files, with LCA storing all the data in one JSON file and SBT opting for saving a description of the index structure in JSON, and all the data into a hidden directory with many files.

We distribute some prepared databases (with SBT indices) for Genbank and RefSeq as compressed TAR files. The compressed file is ~8GB, but after decompressing it turns into almost 200k files in a hidden directory, using about 40 GB of disk space.

Can we avoid generating so many hidden files?

The initial issue in this saga is dib-lab/sourmash#490, and the idea was to take the existing support for multiple data storages (hidden dir, TAR files, IPFS and Redis) and save the index description in the storage, allowing loading everything from the storage. Since we already had the databases as TAR files, the first test tried to use them but it didn't take long to see it was a doomed approach: TAR files are terrible from random access (or at least the tarfile module in Python is).

Zip files showed up as a better alternative, and it helps that Python has the zipfile module already available in the standard library. Initial tests were promising, and led to dib-lab/sourmash#648. The main issue was performance: compressing and decompressing was slow, but there was also another limitation...

Loading Nodegraphs from a memory buffer

Another challenge was efficiently loading the data from a storage. The two core methods in a storage are save(location, content), where content is a bytes buffer, and load(location), which returns a bytes buffer that was previously saved. This didn't interact well with the khmer Nodegraphs (the Bloom Filter we use for SBTs), since khmer only loads data from files, not from memory buffers. We ended up doing a temporary file dance, which made things slower for the default storage (hidden dir), where it could have been optimized to work directly with files, and involved interacting with the filesystem for the other storages (IPFS and Redis could be pulling data directly from the network, for example).

This one could be fixed in khmer by exposing C++ stream methods, and I did a small PoC to test the idea. While doable, this is something that was happening while the sourmash conversion to Rust was underway, and depending on khmer was a problem for my Webassembly aspirations... so, having the Nodegraph implemented in Rust seemed like a better direction, That has actually been quietly living in the sourmash codebase for quite some time, but it was never exposed to the Python (and it was also lacking more extensive tests).

After the release of sourmash 3 and the replacement of the C++ for the Rust implementation, all the pieces for exposing the Nodegraph where in place, so dib-lab/sourmash#799 was the next step. It wasn't a priority at first because other optimizations (that were released in 3.1 and 3.2) were more important, but then it was time to check how this would perform. And...

Your Rust code is not so fast, huh?

Turns out that my Nodegraph loading code was way slower than khmer. The Nodegraph binary format is well documented, and doing an initial implementation wasn't so hard by using the byteorder crate to read binary data with the right endianess, and then setting the appropriate bits in the internal fixedbitset in memory. But the khmer code doesn't parse bit by bit: it reads a long char buffer directly, and that is many orders of magnitude faster than setting bit by bit.

And there was no way to replicate this behavior directly with fixedbitset. At this point I could either bit-indexing into a large buffer and lose all the useful methods that fixedbitset provides, or try to find a way to support loading the data directly into fixedbitset and open a PR.

I chose the PR (and even got #42! =]).

It was more straightforward than I expected, but it did expose the internal representation of fixedbitset, so I was a bit nervous it wasn't going to be merged. But bluss was super nice, and his suggestions made the PR way better! This simplified the final Nodegraph code, and actually was more correct (because I was messing a few corner cases when doing the bit-by-bit parsing before). Win-win!

Nodegraphs are kind of large, can we compress them?

Being able to save and load Nodegraphs in Rust allowed using memory buffers, but also opened the way to support other operations not supported in khmer Nodegraphs. One example is loading/saving compressed files, which is supported for Countgraph (another khmer data structure, based on Count-Min Sketch) but not in Nodegraph.

If only there was an easy way to support working with compressed files...

Oh wait, there is! niffler is a crate that I made with Pierre Marijon based on some functionality I saw in one of his projects, and we iterated a bit on the API and documented everything to make it more useful for a larger audience. niffler tries to be as transparent as possible, with very little boilerplate when using it but with useful features nonetheless (like auto detection of the compression format). If you want more about the motivation and how it happened, check this Twitter thread.

The cool thing is that adding compressed files support in sourmash was mostly one-line changes for loading (and a bit more for saving, but mostly because converting compression levels could use some refactoring).

Putting it all together: zipped SBT indices

With all these other pieces in places, it's time to go back to dib-lab/sourmash#648. Compressing and decompressing with the Python zipfile module is slow, but Zip files can also be used just for storage, handing back the data without extracting it. And since we have compression/decompression implemented in Rust with niffler, that's what the zipped sourmash databases are: data is loaded and saved into the Zip file without using the Python module compression/decompression, and all the work is done before (or after) in the Rust side.

This allows keeping the Zip file with similar sizes to the original TAR files we started with, but with very low overhead for decompression. For compression we opted for using Gzip level 1, which doesn't compress perfectly but also doesn't take much longer to run:

In this table, 0 is without compression, while 9 is the best compression. The size difference from 1 to 9 is only 6 MB (~2% difference) but runs 5x faster, and it's only 30% slower than saving the uncompressed data.

The last challenge was updating an existing Zip file. It's easy to support appending new data, but if any of the already existing data in the file changes (which happens when internal nodes change in the SBT, after a new dataset is inserted) then there is no easy way to replace the data in the Zip file. Worse, the Python zipfile will add the new data while keeping the old one around, leading to ginormous files over time¹ So, what to do?

I ended up opting for dealing with the complexity and complicating the ZipStorage implementation a bit, by keeping a buffer for new data. If it's a new file or it already exists but there are no insertions the buffer is ignored and all works as before.

If the file exists and new data is inserted, then it is first stored in the buffer (where it might also replace a previous entry with the same name). In this case we also need to check the buffer when trying to load some data (because it might exist only in the buffer, and not in the original file).

Finally, when the ZipStorage is closed it needs to verify if there are new items in the buffer. If not, it is safe just to close the original file. If there are new items but they were not present in the original file, then we can append the new data to the original file. The final case is if there are new items that were also in the original file, and in this case a new Zip file is created and all the content from buffer and original file are copied to it, prioritizing items from the buffer. The original file is replaced by the new Zip file.

Turns out this worked quite well! And so the PR was merged =]

The future

Zipped databases open the possibility of distributing extra data that might be useful for some kinds of analysis. One thing we are already considering is adding taxonomy information, let's see what else shows up.

Having Nodegraph in Rust is also pretty exciting, because now we can change the internal representation for something that uses less memory (maybe using RRR encoding?), but more importantly: now they can also be used with Webassembly, which opens many possibilities for running not only signature computation but also search and gather in the browser, since now we have all the pieces to build it.

Comments?

• Thread on Twitter

Footnotes

1. The zipfile module does throw a UserWarning pointing that duplicated files were inserted, which is useful during development but generally doesn't show during regular usage... ↩

Em novembro passado, colaboradores latinoamericanos do KDE desembarcaram em Salvador/Brasil para participarem de mais uma edição do LaKademy – o Latin American Akademy. Aquela foi a sétima edição do evento (ou oitava, se você contar o Akademy-BR como o primeiro LaKademy) e a segunda com Salvador como a cidade que hospedou o evento. Sem problemas para mim: na verdade, adoraria me mudar e viver ao menos alguns anos em Salvador, cidade que gosto baste.

Foto em grupo do LaKademy 2019

Minhas principais tarefas durante o evento foram em 2 projetos: Cantor e Sprat, um “editor de rascunhos de artigos acadêmicos”. Além deles, ajudei também com tarefas de promoção como o site do LaKademy.

Nos trabalhos sobre o Cantor me foquei naqueles relacionados com organização. Por exemplo, pedi aos sysadmins que migrassem o repositório para o Gitlab do KDE e crei um site específico para o Cantor em cantor.kde.org usando o novo template em Jekyll para projetos do KDE.

O novo site é uma boa adição ao Cantor porque nós queremos comunicar melhor e mais diretamente com nossa comunidade de usuários. O site tem um blog próprio e uma seção de changelog para tornar mais fácil à comunidde seguir as notícias e principais mudanças no software.

A migração para o Gitlab nos permite utilizar o Gitlab CI como uma alternativa para integração contínua no Cantor. Eu orientei o trabalho do Rafael Gomes (que ainda não teve merge) para termos isso disponível pro projeto.

Além dos trabalhos no Cantor, desenvolvi algumas atividades relacionadas ao Sprat, um editor de rascunhos de artigos científicos em inglês. Este softwar usa katepart para implementar e metodologia de escrita de artigos científicos em inglês conhecida como PROMETHEUS, conforme descrita neste livro, como uma tentativa de auxiliar estudantes e pesquisadores em geral na tarefa de escrever artigos científicos. Durante o LaKademy finalizei o port para Qt5 e, tomara, espero lançar o projeto este ano.

Nas atividades mais sociais, participei da famosíssima reunião de promo, que discute as ações futuras do KDE para a América Latina. Nossa principal decisão foi organizar e participar mais de  eventos pequenos e distribuídos em várias cidades, marcando a presença do KDE em eventos consolidados como o FLISoL e o Software Freedom Day, e mais – mas agora, em tempos de COVID-19, isso não é mais viável. Outra decisão foi mover a organização do KDE Brasil do Phabricator para o Gitlab.

Contribuidores do KDE trabalhando pesado

Para além da parte técnica, este LaKademy foi uma oportunidade para encontrar velhos e novos amigos, beber algumas cervejas, saborear a maravilhosa cozinha bahiana, e se divertir entre um commit e outro.

Gostaria de agradecer ao KDE e.V. por apoiar o LaKademy, e Caio e Icaro por terem organizado essa edição do evento. Não vejo a hora de participar do próximo LaKademy e que isso seja o mais rápido possível!

Foto em grupo do Akademy 2019

Em setembro de 2019 a cidade italiana de Milão sediou o principal encontro mundial dos colaboradores do KDE – o Akademy, onde membros de diferentes áreas como tradutores, desenvolvedores, artistas, pessoal de promo e mais se reúnem por alguns dias para pensar e construir o futuro dos projetos e comunidade(s) do KDE

Antes de chegar ao Akademy tomei um voo do Brasil para Portugal a fim de participar do EPIA, uma conferência sobre inteligência artificial que aconteceu na pequena cidade de Vila Real, região do Porto. Após essa atividade acadêmica, voei do Porto à Milão e iniciei minha participação no Akademy 2019.

Infelizmente pousei no final da primeira manhã do evento, o que me fez perder apresentações interessantes sobre Qt 6 e os novos KDE’ Goals. Pela tarde pude algumas palestras sobre temas que também me chamam atenção, como Plasma para dispositivos móveis, MyCroft para a indústria automotiva, relatório do KDE e.V. e um showcase de estudantes do Google Summer of Code e do Season of KDE – muito legal ver projetos incríveis desenvolvidos pelos novatos.

No segundo dia me chamou atenção as palestras sobre o KPublicTransportation, LibreOffice para o Plasma, Get Hot New Stuffs – que imagino utilizarei em um futuro projeto – e a apresentação do Caio sobre o kpmcore.

Após a festa do evento (comentários sobre ela apenas pessoalmente), os dias seguintes foram preenchidos pelos BoFs, para mim a parte mais interessante do Akademy.

O workshop do Gitlab foi interessante porque pudemos discutir temas específicos sobre a migração do KDE para esta ferramenta. Estou adorando esse movimento e espero que todos os projetos do KDE façam essa migração o quanto antes. Cantor já está por lá há algum tempo.

No BoF do KDE websites, pude entender um pouco melhor sobre o novo tema do Jekyll utilizado pelos nossos sites. Em adição, aguardo que logo mais possamos aplicar internacionalização nessas páginas, tornando-as traduzíveis para quaisquer idiomas. Após participar e tomar informações nesse evento, criei um novo website pro Cantor durante o LaKademy 2019.

O BoF do KDE Craft foi interessante para ver como compilar e distribuir nosso software na loja do Windows (pois é, vejam só o que estou escrevendo…). Espero trabalhar com esse tema durante o ano de forma a disponibilizar um pacote do Cantor naquela loja ((pois é)²).

Também participei do workshop sobre QML e Kirigami realizado pelo pessoal do projeto Maui. Kirigami é algo que tenho mantido o olho para futuros projetos.

Finalmente, participei do BoF “All About the Apps Kick Off”. Pessoalmente, penso que esse é o futuro do KDE: uma comunidade internacional que produz software livre de alta qualidade e seguro para diferentes plataformas, do desktop ao mobile. De fato, isto é como o KDE está atualmente organizado e funcionando, mas nós não conseguimos comunicar isso muito bem para o público. Talvez, com as mudanças em nossa forma de lançamentos, somados aos websites para projetos específicos e distribuição em diferentes lojas de aplicativos, possamos mudar a maneira como o público vê a nossa comunidade.

O day trip do Akademy 2019 foi no lago Como, na cidade de Varenna. Uma viagem linda, passei o tempo todo imaginando que lá poderia ser um bom lugar para eu passar uma lua de mel :D. Espero voltar lá no futuro próximo, e passar alguns dias viajando entre cidades como ela.

Eu em Varenna

Gostaria de agradecer a todo o time local, Riccardo e seus amigos, por organizarem essa edição incrível do Akademy. Milão é uma cidade muito bonita, com comida deliciosa (carbonara!), lugares históricos para visitar e descobrir mais sobre os italianos e a sofisticada capital da alta moda.

Finalmente, meus agradecimentos ao KDE e.V. por patrocinar minha participação no Akademy.

Neste link estão disponíveis vídeos das apresentações e BoFs do Akademy 2019.

Fala pessoal, tudo bom?

Nos vídeo abaixo vou mostrar como podemos configurar um CI de uma aplicação Django usando Github Actions.

https://www.youtube.com/watch?v=KpSlY8leYFY.

sourmash 3 was released last week, finally landing the Rust backend. But, what changes when developing new features in sourmash? I was thinking about how to best document this process, and since PR #826 is a short example touching all the layers I decided to do a small walkthrough.

Shall we?

The problem

The first step is describing the problem, and trying to convince reviewers (and yourself) that the changes bring enough benefits to justify a merge. This is the description I put in the PR:

Calling .add_hash() on a MinHash sketch is fine, but if you're calling it all the time it's better to pass a list of hashes and call .add_many() instead. Before this PR add_many just called add_hash for each hash it was passed, but now it will pass the full list to Rust (and that's way faster).

No changes for public APIs, and I changed the _signatures method in LCA to accumulate hashes for each sig first, and then set them all at once. This is way faster, but might use more intermediate memory (I'll evaluate this now).

There are many details that sound like jargon for someone not familiar with the codebase, but if I write something too long I'll probably be wasting the reviewers time too. The benefit of a very detailed description is extending the knowledge for other people (not necessarily the maintainers), but that also takes effort that might be better allocated to solve other problems. Or, more realistically, putting out other fires =P

Nonetheless, some points I like to add in PR descriptions: - why is there a problem with the current approach? - is this the minimal viable change, or is it trying to change too many things at once? The former is way better, in general. - what are the trade-offs? This PR is using more memory to lower the runtime, but I hadn't measure it yet when I opened it. - Not changing public APIs is always good to convince reviewers. If the project follows a semantic versioning scheme, changes to the public APIs are major version bumps, and that can brings other consequences for users.

Setting up for changing code

If this was a bug fix PR, the first thing I would do is write a new test triggering the bug, and then proceed to fix it in the code (Hmm, maybe that would be another good walkthrough?). But this PR is making performance claims ("it's going to be faster"), and that's a bit hard to codify in tests. ¹ Since it's also proposing to change a method (_signatures in LCA indices) that is better to benchmark with a real index (and not a toy example), I used the same data and command I run in sourmash_resources to check how memory consumption and runtime changed. For reference, this is the command:

sourmash search -o out.csv --scaled 2000 -k 51 HSMA33OT.fastq.gz.sig genbank-k51.lca.json.gz

I'm using the benchmark feature from snakemake in sourmash_resources to track how much memory, runtime and I/O is used for each command (and version) of sourmash, and generate the plots in the README in that repo. That is fine for a high-level view ("what's the maximum memory used?"), but not so useful for digging into details ("what method is consuming most memory?").

Another additional problem is the dual² language nature of sourmash, where we have Python calling into Rust code (via CFFI). There are great tools for measuring and profiling Python code, but they tend to not work with extension code...

So, let's bring two of my favorite tools to help!

Memory profiling: heaptrack

heaptrack is a heap profiler, and I first heard about it from Vincent Prouillet. Its main advantage over other solutions (like valgrind's massif) is the low overhead and... how easy it is to use: just stick heaptrack in front of your command, and you're good to go!

Example output:

$ heaptrack sourmash search -o out.csv --scaled 2000 -k 51 HSMA33OT.fastq.gz.sig genbank-k51.lca.json.gz

heaptrack stats:
        allocations:            1379353
        leaked allocations:     1660
        temporary allocations:  168984
Heaptrack finished! Now run the following to investigate the data:

  heaptrack --analyze heaptrack.sourmash.66565.gz

heaptrack --analyze is a very nice graphical interface for analyzing the results, but for this PR I'm mostly focusing on the Summary page (and overall memory consumption). Tracking allocations in Python doesn't give many details, because it shows the CPython functions being called, but the ability to track into the extension code (Rust) allocations is amazing for finding bottlenecks (and memory leaks =P). ³

CPU profiling: py-spy

Just as other solutions exist for profiling memory, there are many for profiling CPU usage in Python, including profile and cProfile in the standard library. Again, the issue is being able to analyze extension code, and bringing the cannon (the perf command in Linux, for example) looses the benefit of tracking Python code properly (because we get back the CPython functions, not what you defined in your Python code).

Enters py-spy by Ben Frederickson, based on the rbspy project by Julia Evans. Both use a great idea: read the process maps for the interpreters and resolve the full stack trace information, with low overhead (because it uses sampling). py-spy also goes further and resolves native Python extensions stack traces, meaning we can get the complete picture all the way from the Python CLI to the Rust core library!⁴

py-spy is also easy to use: stick py-spy record --output search.svg -n -- in front of the command, and it will generate a flamegraph in search.svg. The full command for this PR is

py-spy record --output search.svg -n -- sourmash search -o out.csv --scaled 2000 -k 51 HSMA.fastq.sig genbank-k51.lca.json.gz

Show me the code!

OK, OK, sheesh. But it's worth repeating: the code is important, but there are many other aspects that are just as important =]

Replacing add_hash calls with one add_many

Let's start at the _signatures() method on LCA indices. This is the original method:

@cached_property
def _signatures(self):
    "Create a _signatures member dictionary that contains {idx: minhash}."
    from .. import MinHash

    minhash = MinHash(n=0, ksize=self.ksize, scaled=self.scaled)

    debug('creating signatures for LCA DB...')
    sigd = defaultdict(minhash.copy_and_clear)

    for (k, v) in self.hashval_to_idx.items():
        for vv in v:
            sigd[vv].add_hash(k)

    debug('=> {} signatures!', len(sigd))
    return sigd

sigd[vv].add_hash(k) is the culprit. Each call to .add_hash has to go thru CFFI to reach the extension code, and the overhead is significant. It is a similar situation to accessing array elements in NumPy: it works, but it is way slower than using operations that avoid crossing from Python to the extension code. What we want to do instead is call .add_many(hashes), which takes a list of hashes and process it entirely in Rust (ideally. We will get there).

But, to have a list of hashes, there is another issue with this code.

for (k, v) in self.hashval_to_idx.items():
    for vv in v:
        sigd[vv].add_hash(k)

There are two nested for loops, and add_hash is being called with values from the inner loop. So... we don't have the list of hashes beforehand.

But we can change the code a bit to save the hashes for each signature in a temporary list, and then call add_many on the temporary list. Like this:

temp_vals = defaultdict(list)

for (k, v) in self.hashval_to_idx.items():
    for vv in v:
        temp_vals[vv].append(k)

for sig, vals in temp_vals.items():
    sigd[sig].add_many(vals)

There is a trade-off here: if we save the hashes in temporary lists, will the memory consumption be so high that the runtime gains of calling add_many in these temporary lists be cancelled?

Time to measure it =]

Wait, it got worse?!?! Building temporary lists only takes time and memory, and bring no benefits!

This mystery goes away when you look at the add_many method:

def add_many(self, hashes):
    "Add many hashes in at once."
    if isinstance(hashes, MinHash):
        self._methodcall(lib.kmerminhash_add_from, hashes._objptr)
    else:
        for hash in hashes:
            self._methodcall(lib.kmerminhash_add_hash, hash)

The first check in the if statement is a shortcut for adding hashes from another MinHash, so let's focus on else part... And turns out that add_many is lying! It doesn't process the hashes in the Rust extension, but just loops and call add_hash for each hash in the list. That's not going to be any faster than what we were doing in _signatures.

Time to fix add_many!

Oxidizing add_many

The idea is to change this loop in add_many:

for hash in hashes:
    self._methodcall(lib.kmerminhash_add_hash, hash)

with a call to a Rust extension function:

self._methodcall(lib.kmerminhash_add_many, list(hashes), len(hashes))

self._methodcall is a convenience method defined in RustObject which translates a method-like call into a function call, since our C layer only has functions. This is the C prototype for this function:

void kmerminhash_add_many(
    KmerMinHash *ptr,
    const uint64_t *hashes_ptr,
    uintptr_t insize
  );

You can almost read it as a Python method declaration, where KmerMinHash *ptr means the same as the self in Python methods. The other two arguments are a common idiom when passing pointers to data in C, with insize being how many elements we have in the list. ⁵. CFFI is very good at converting Python lists into pointers of a specific type, as long as the type is of a primitive type (uint64_t in our case, since each hash is a 64-bit unsigned integer number).

And the Rust code with the implementation of the function:

ffi_fn! {
unsafe fn kmerminhash_add_many(
    ptr: *mut KmerMinHash,
    hashes_ptr: *const u64,
    insize: usize,
  ) -> Result<()> {
    let mh = {
        assert!(!ptr.is_null());
        &mut *ptr
    };

    let hashes = {
        assert!(!hashes_ptr.is_null());
        slice::from_raw_parts(hashes_ptr as *mut u64, insize)
    };

    for hash in hashes {
      mh.add_hash(*hash);
    }

    Ok(())
}
}

Let's break what's happening here into smaller pieces. Starting with the function signature:

ffi_fn! {
unsafe fn kmerminhash_add_many(
    ptr: *mut KmerMinHash,
    hashes_ptr: *const u64,
    insize: usize,
  ) -> Result<()>

The weird ffi_fn! {} syntax around the function is a macro in Rust: it changes the final generated code to convert the return value (Result<()>) into something that is valid C code (in this case, void). What happens if there is an error, then? The Rust extension has code for passing back an error code and message to Python, as well as capturing panics (when things go horrible bad and the program can't recover) in a way that Python can then deal with (raising exceptions and cleaning up). It also sets the #[no_mangle] attribute in the function, meaning that the final name of the function will follow C semantics (instead of Rust semantics), and can be called more easily from C and other languages. This ffi_fn! macro comes from symbolic, a big influence on the design of the Python/Rust bridge in sourmash.

unsafe is the keyword in Rust to disable some checks in the code to allow potentially dangerous things (like dereferencing a pointer), and it is required to interact with C code. unsafe doesn't mean that the code is always unsafe to use: it's up to whoever is calling this to verify that valid data is being passed and invariants are being preserved.

If we remove the ffi_fn! macro and the unsafe keyword, we have

fn kmerminhash_add_many(
    ptr: *mut KmerMinHash,
    hashes_ptr: *const u64,
    insize: usize
  );

At this point we can pretty much map between Rust and the C function prototype:

void kmerminhash_add_many(
    KmerMinHash *ptr,
    const uint64_t *hashes_ptr,
    uintptr_t insize
  );

Some interesting points:

• We use fn to declare a function in Rust.
• The type of an argument comes after the name of the argument in Rust, while it's the other way around in C. Same for the return type (it is omitted in the Rust function, which means it is -> (), equivalent to a void return type in C).
• In Rust everything is immutable by default, so we need to say that we want a mutable pointer to a KmerMinHash item: *mut KmerMinHash). In C everything is mutable by default.
• u64 in Rust -> uint64_t in C
• usize in Rust -> uintptr_t in C

Let's check the implementation of the function now. We start by converting the ptr argument (a raw pointer to a KmerMinHash struct) into a regular Rust struct:

let mh = {
    assert!(!ptr.is_null());
    &mut *ptr
};

This block is asserting that ptr is not a null pointer, and if so it dereferences it and store in a mutable reference. If it was a null pointer the assert! would panic (which might sound extreme, but is way better than continue running because dereferencing a null pointer is BAD). Note that functions always need all the types in arguments and return values, but for variables in the body of the function Rust can figure out types most of the time, so no need to specify them.

The next block prepares our list of hashes for use:

let hashes = {
    assert!(!hashes_ptr.is_null());
    slice::from_raw_parts(hashes_ptr as *mut u64, insize)
};

We are again asserting that the hashes_ptr is not a null pointer, but instead of dereferencing the pointer like before we use it to create a slice, a dynamically-sized view into a contiguous sequence. The list we got from Python is a contiguous sequence of size insize, and the slice::from_raw_parts function creates a slice from a pointer to data and a size.

Oh, and can you spot the bug? I created the slice using *mut u64, but the data is declared as *const u64. Because we are in an unsafe block Rust let me change the mutability, but I shouldn't be doing that, since we don't need to mutate the slice. Oops.

Finally, let's add hashes to our MinHash! We need a for loop, and call add_hash for each hash:

for hash in hashes {
  mh.add_hash(*hash);
}

Ok(())

We finish the function with Ok(()) to indicate no errors occurred.

Why is calling add_hash here faster than what we were doing before in Python? Rust can optimize these calls and generate very efficient native code, while Python is an interpreted language and most of the time don't have the same guarantees that Rust can leverage to generate the code. And, again, calling add_hash here doesn't need to cross FFI boundaries or, in fact, do any dynamic evaluation during runtime, because it is all statically analyzed during compilation.

Putting it all together

And... that's the PR code. There are some other unrelated changes that should have been in new PRs, but since they were so small it would be more work than necessary. OK, that's a lame excuse: it's confusing for reviewers to see these changes here, so avoid doing that if possible!

But, did it work?

We are using 200 MB of extra memory, but taking less than half the time it was taking before. I think this is a good trade-off, and so did the reviewer and the PR was approved.

Hopefully this was useful, 'til next time!

Comments?

• Thread on Mastodon
• Thread on Twitter
• Lobste.rs submission

Bonus: list or set?

The first version of the PR used a set instead of a list to accumulate hashes. Since a set doesn't have repeated elements, this could potentially use less memory. The code:

temp_vals = defaultdict(set)

for (k, v) in self.hashval_to_idx.items():
    for vv in v:
        temp_vals[vv].add(k)

for sig, vals in temp_vals.items():
    sigd[sig].add_many(vals)

The runtime was again half of the original, but...

... memory consumption was almost 2.5 times the original! WAT

The culprit this time? The new kmerminhash_add_many call in the add_many method. This one:

self._methodcall(lib.kmerminhash_add_many, list(hashes), len(hashes))

CFFI doesn't know how to convert a set into something that C understands, so we need to call list(hashes) to convert it into a list. Since Python (and CFFI) can't know if the data is going to be used later ⁶ it needs to keep it around (and be eventually deallocated by the garbage collector). And that's how we get at least double the memory being allocated...

There is another lesson here. If we look at the for loop again:

for (k, v) in self.hashval_to_idx.items():
    for vv in v:
        temp_vals[vv].add(k)

each k is already unique because they are keys in the hashval_to_idx dictionary, so the initial assumption (that a set might save memory because it doesn't have repeated elements) is... irrelevant for the problem =]

Footnotes

1. We do have https://asv.readthedocs.io/ set up for micro-benchmarks, and now that I think about it... I could have started by writing a benchmark for add_many, and then showing that it is faster. I will add this approach to the sourmash PR checklist =] ↩

2. or triple, if you count C ↩

3. It would be super cool to have the unwinding code from py-spy in heaptrack, and be able to see exactly what Python methods/lines of code were calling the Rust parts... ↩

4. Even if py-spy doesn't talk explicitly about Rust, it works very very well, woohoo! ↩

5. Let's not talk about lack of array bounds checks in C... ↩

6. something that the memory ownership model in Rust does, BTW ↩

What is Tsuru?

What is Tsuru public cloud? What does "beta availability" means?

Which development platforms are going to be available?

How limited is it going to be?

When will it be available?

What if I don't want to wait?

% curl -O http://nginx.org/download/nginx-1.0.9.tar.gz
% tar -xzf nginx-1.0.9.tar.gz

% ./configure --prefix=/opt/nginx --user=nginx --group=nginx
% make
% [sudo] make install

% [sudo] adduser --system --no-create-home --disabled-login --disabled-password --group nginx

% [sudo] /opt/nginx/sbin/nginx

Linode provides an init script that uses start-stop-daemon, you might want to use it.

nginx configuration

• nginx should start workers with the nginx user
• nginx should have two worker processes
• the PID should be stored in the /opt/nginx/log/nginx.pid file
• nginx must have an access log in /opt/nginx/logs/access.log
• the configuration for the Django project we’re going to develop should be versioned with the entire code, so it must be included in the nginx.conf file (assume that the library project is in the directory /opt/projects).

user  nginx;
worker_processes  2;

pid logs/nginx.pid;

events {
    worker_connections  1024;
}

http {
    include       mime.types;
    default_type  application/octet-stream;

    log_format  main  '$remote_addr - $remote_user [$time_local] "$request" '
                     '$status $body_bytes_sent "$http_referer" '
                     '"$http_user_agent" "$http_x_forwarded_for"';

    access_log  logs/access.log  main;

    sendfile           on;
    keepalive_timeout  65;

    include /opt/projects/showcase/nginx.conf;
}

server {
    listen 80;
    server_name localhost;

    charset utf-8;

    location / {
        proxy_set_header    X-Real-IP   $remote_addr;
        proxy_set_header    Host        $http_host;
        proxy_set_header    X-Forwarded-For $proxy_add_x_forwarded_for;

        proxy_pass http://localhost:8000;
    }

    location /static {
        root /opt/projects/showcase/;
        expires 1d;
    }
}

• X-Real-IP: forwards the remote address to the upstream server, so it can know the real IP of the user. When nginx forwards the request to gunicorn, without this header, all gunicorn will know is that there is a request coming from localhost (or wherever the nginx server is), the remote address is always the IP address of the machine where nginx is running (who actually make the request to gunicorn)
• Host: the Host header is forwarded so gunicorn can treat different requests for different hosts. Without this header, it will be impossible to Gunicorn to have these constraints
• X-Forwarded-For: also known as XFF, this header provide more precise information about the real IP who makes the request. Imagine there are 10 proxies between the user machine and your webserver, the XFF header will all these proxies comma separated. In order to not turn a proxy into an anonymizer, it’s a good practice to always forward this header.

Python
======

--------------

If
==

* Please don't use ()
* Never forget the ``:`` at the end of the line

Check this code:

.. sourcecode:: python

    x, y = 1, 2
    if x > y:
        print 'x is greater'

--------------

For
===

* ``for`` iterates over a sequence
* Never forget the ``:`` at the end of the line

Check this code:

.. sourcecode:: python

    numbers = [1, 2, 3, 4, 5,]
    for number in numbers:
        print number

--------------

While
=====

* ``while`` is like ``if``, but executes while the codition is ``True``
* please don't use ()
* never forget the ``:`` at the end of the line

Check this code:

.. sourcecode:: python

    from random import randint

    args = (1, 10,)
    x = randint(*args)
    while x != 6:
        x = randint(*args)

--------------

Thank you!
==========

% landslide python.rst

% [sudo] pip install landslide

% landslide --help

What is a splinter sprint?

How to contribute

1. Choose an issue
2. Create a fork
3. Send a pull request

Preparing for the sprint

.
├── SpecRunner.html
├── lib
│   ├── jasmine-1.0.2
│   │   ├── MIT.LICENSE
│   │   ├── jasmine-html.js
│   │   ├── jasmine.css
│   │   └── jasmine.js
│   └── jquery-1.6.1.min.js
├── spec
│   └── ZebrafySpec.js
└── src
    └── jquery.zebrafy.js

describe('Zebrafy', function () {

});

it('should apply classes zebrafy-odd and zebrafy-even to each other table lines', function () {
    var table = $("#zebra-table");
    table.zebrafy();
    expect(table).toBeZebrafyied();
});

beforeEach(function () {
    $('<table id="zebra-table"></table>').appendTo('body');
    for (var i=0; i < 10; i++) {
        $('<tr></tr>').append('<td></td>').append('<td></td>').append('<td></td>').appendTo('#zebra-table');
    };
});

afterEach(function () {
    $("#zebra-table").remove();
});

it('should apply classes zebrafy-odd and zebrafy-even to each other table lines', function () {
    var table = $("#zebra-table");
    table.zebrafy();
    expect(table).toBeZebrafyied();
});

beforeEach(function () {
    this.addMatchers({
        toBeZebrafyied: function() {
            var isZebrafyied = true;

            this.actual.find("tr:even").each(function (index, tr) {
                isZebrafyied = $(tr).hasClass('zebrafy-odd') === false && $(tr).hasClass('zebrafy-even');
                if (!isZebrafyied) {
                    return;
                };
            });

            this.actual.find("tr:odd").each(function (index, tr) {
                isZebrafyied = $(tr).hasClass('zebrafy-odd') && $(tr).hasClass('zebrafy-even') === false;
                if (!isZebrafyied) {
                    return;
                };
            });

            return isZebrafyied;
        }
    });
});

(function ($) {
    $.fn.zebrafy = function () {
        this.find("tr:even").addClass("zebrafy-even");
        this.find("tr:odd").addClass("zebrafy-odd");
    };
})(jQuery);

it('zebrafy should be chainable', function() {
    var table = $("#zebra-table");
    table.zebrafy().addClass('black-bg');
    expect(table.hasClass('black-bg')).toBeTruthy();
});

(function ($) {
    $.fn.zebrafy = function () {
        return this.each(function (index, table) {
            $(table).find("tr:even").addClass("zebrafy-even");
            $(table).find("tr:odd").addClass("zebrafy-odd");
        });
    };
})(jQuery);

browser = Browser("firefox")

browser.visit("https://www.facebook.com")

def do_login_if_need(self, username, password):
    if self.browser.is_element_present_by_css('div.menu_login_container'):
        self.browser.fill('email', username)
        self.browser.fill('pass', password)
        self.browser.find_by_css('div.menu_login_container input[type="submit"]').first.click()
        assert self.browser.is_element_present_by_css('li#navAccount')

1. On Facebook homepage, click on “Events” link, of left menu
2. The “Events” page will load, so click on “Create an Event” button
3. The user see a page with a form to create an event
4. Fill the date and chose the time
5. Define what is the name of the event, where it will happen and write a short description for it
6. Invite some guests
7. Upload a picture for the event
8. Click on “Create Event” button

browser.find_by_css('li#navItem_events a').first.click()

browser.fill('event_startIntlDisplay', '5/21/2011')
browser.select('start_time_min', '480')
browser.fill('name', 'Splinter sprint')
browser.fill('location', 'Rio de Janeiro, Brazil')
browser.fill('desc', 'For more info, check out the #cobratem channel on freenode!')

<select name="gender">
    <option value="m">Male</option>
    <option value="f">Female</option>
</select>

browser.select("gender", "m")

browser.find_by_css('div.eventEditUpload a.uiButton').first.click()

if not browser.is_element_present_by_css('iframe#upload_pic_frame', wait_time=10):
fail("The upload pic iframe did'n't appear :(")

Important: fail is a pseudocode sample and doesn’t exist (if you’re using unittest library, you can invoke self.fail in a TestCase, exactly what I did in complete snippet for this example, available at Github).

with browser.get_iframe('upload_pic_frame') as frame:
    frame.attach_file('pic', picture_path)
    time.sleep(10)

browser.find_by_css('label.uiButton input[type="submit"]').first.click()
title = browser.find_by_css('h1 span').first.text
assert title == 'Splinter sprint'

nginx

% ./configure --prefix=/opt/nginx --user=nginx --group=nginx --with-http_ssl_module --add-module={your memc_module source path}
% make
% sudo make install

user  nginx;
worker_processes  1;

error_log  logs/error.log;

events {
    worker_connections  1024;
}

http {
    include       mime.types;
    default_type  application/octet-stream;

    log_format  main  '$remote_addr - $remote_user [$time_local] "$request" '
                                '$status $body_bytes_sent "$http_referer" '
                                '"$http_user_agent" "$http_x_forwarded_for"';

    access_log  logs/access.log  main;

    sendfile        on;
    #tcp_nopush     on;

    #keepalive_timeout  0;
    keepalive_timeout  65;

    #gzip  on;

    include /opt/nginx/sites-enabled/*;
}

server {
    listen       80;
    server_name  localhost;

    default_type  text/html;

    location / {
        proxy_set_header    X-Real-IP   $remote_addr;
        proxy_set_header    Host        $http_host;
        proxy_set_header    X-Forwarded-For $proxy_add_x_forwarded_for;

        if ($request_method = POST) {
            proxy_pass      http://localhost:8080;
            break;
        }

        set $memcached_key   "$uri";
        memcached_pass      127.0.0.1:11211;

        error_page  501 404 502 = /fallback$uri;
    }

    location /fallback/ {
        internal;    

        proxy_set_header    X-Real-IP   $remote_addr;
        proxy_set_header    Host        $http_host;
        proxy_set_header    X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_redirect      off;

        proxy_pass          http://localhost:8080;
    }

}

• 404: memcached module returns a 404 error when the key is not on memcached server;
• 502: memcached module returns a 502 error when it can’t found memcached server.

Java application

package com.franciscosouza.memcached.filter;

import java.io.IOException;
import java.io.PrintWriter;
import java.io.StringWriter;
import java.net.InetSocketAddress;

import javax.servlet.Filter;
import javax.servlet.FilterChain;
import javax.servlet.FilterConfig;
import javax.servlet.ServletException;
import javax.servlet.ServletOutputStream;
import javax.servlet.ServletRequest;
import javax.servlet.ServletResponse;
import javax.servlet.http.HttpServletRequest;
import javax.servlet.http.HttpServletResponse;
import javax.servlet.http.HttpServletResponseWrapper;

import net.spy.memcached.MemcachedClient;

/**
 * Servlet Filter implementation class MemcachedFilter
 */
public class MemcachedFilter implements Filter {

    private MemcachedClient mmc;

    static class MemcachedHttpServletResponseWrapper extends HttpServletResponseWrapper {

        private StringWriter sw = new StringWriter();

        public MemcachedHttpServletResponseWrapper(HttpServletResponse response) {
            super(response);
        }

        public PrintWriter getWriter() throws IOException {
            return new PrintWriter(sw);
        }

        public ServletOutputStream getOutputStream() throws IOException {
            throw new UnsupportedOperationException();
        }

        public String toString() {
            return sw.toString();
        }
    }

    /**
     * Default constructor.
     */
    public MemcachedFilter() {
    }

    /**
     * @see Filter#destroy()
     */
    public void destroy() {
    }

    /**
     * @see Filter#doFilter(ServletRequest, ServletResponse, FilterChain)
     */
    public void doFilter(ServletRequest request, ServletResponse response, FilterChain chain) throws IOException, ServletException {
        MemcachedHttpServletResponseWrapper wrapper = new MemcachedHttpServletResponseWrapper((HttpServletResponse) response);
        chain.doFilter(request, wrapper);

        HttpServletRequest inRequest = (HttpServletRequest) request;
        HttpServletResponse inResponse = (HttpServletResponse) response;

        String content = wrapper.toString();

        PrintWriter out = inResponse.getWriter();
        out.print(content);

        if (!inRequest.getMethod().equals("POST")) {
            String key = inRequest.getRequestURI();
            mmc.set(key, 5, content);
        }
    }

    /**
     * @see Filter#init(FilterConfig)
     */
    public void init(FilterConfig fConfig) throws ServletException {
        try {
            mmc = new MemcachedClient(new InetSocketAddress("localhost", 11211));
        } catch (IOException e) {
            e.printStackTrace();
            throw new ServletException(e);
        }
    }
}

web.xml

<?xml version="1.0" encoding="UTF-8"?>
<web-app xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns="http://java.sun.com/xml/ns/javaee" xmlns:web="http://java.sun.com/xml/ns/javaee/web-app_2_5.xsd" xsi:schemaLocation="http://java.sun.com/xml/ns/javaee http://java.sun.com/xml/ns/javaee/web-app_2_5.xsd" id="WebApp_ID" version="2.5">
    <display-name>memcached sample</display-name>

    <filter>
        <filter-name>vraptor</filter-name>
        <filter-class>br.com.caelum.vraptor.VRaptor</filter-class>
    </filter>
    
    <filter>
        <filter-name>memcached</filter-name>
        <filter-class>com.franciscosouza.memcached.filter.MemcachedFilter</filter-class>
    </filter>
    
    <filter-mapping>
        <filter-name>memcached</filter-name>
        <url-pattern>/*</url-pattern>
    </filter-mapping>

    <filter-mapping>
        <filter-name>vraptor</filter-name>
        <url-pattern>/*</url-pattern>
        <dispatcher>FORWARD</dispatcher>
        <dispatcher>REQUEST</dispatcher>
    </filter-mapping>

</web-app>

% wget http://www.tipfy.org/tipfy.build.tar.gz
% tar -xvzf tipfy.0.6.2.build.tar.gz
% mv project gaeseries

application: gaeseries
version: 4
runtime: python
api_version: 1

derived_file_type:
- python_precompiled

handlers:
- url: /(robots\.txt|favicon\.ico)
  static_files: static/\1
  upload: static/(.*)

- url: /remote_api
  script: $PYTHON_LIB/google/appengine/ext/remote_api/handler.py
  login: admin

- url: /_ah/queue/deferred
  script: main.py
  login: admin

- url: /.* 
 script: main.py

% mkdir blog
% touch blog/__init__.py

# -*- coding: utf-8 -*-
"""
    config
    ~~~~~~

    Configuration settings.

    :copyright: 2009 by tipfy.org.
    :license: BSD, see LICENSE for more details.
"""
config = {}

# Configurations for the 'tipfy' module.
config['tipfy'] = {
    # Enable debugger. It will be loaded only in development.
    'middleware': [
        'tipfy.ext.debugger.DebuggerMiddleware',
    ],
    # Enable the Hello, World! app example.
    'apps_installed': [
        'apps.hello_world',
        'apps.blog',
    ],
}

from google.appengine.ext import db

class Post(db.Model):
    title = db.StringProperty(required = True)
    content = db.TextProperty(required = True)
    when = db.DateTimeProperty(auto_now_add = True)
    author = db.UserProperty(required = True)

# -*- coding: utf-8 -*-
from tipfy import RequestHandler
from tipfy.ext.jinja2 import render_response
from models import Post

class PostListingHandler(RequestHandler):
    def get(self):
        posts = Post.all()
        return render_response('list_posts.html', posts=posts)

<html>
    <head>
      <meta http-equiv="Content-type" content="text/html; charset=utf-8"/>
      <title>{% block title %}{% endblock %}</title>
    </head>
    <body id="">
        {% block content %}{% endblock %}
    </body>
</html>

{% extends "base.html" %}

{% block title %}
    Posts list
{% endblock %}

{% block content %}
    Listing all posts:

    <ul>
        {% for post in posts %}
            <li>
                {{ post.title }} (written by {{ post.author.nickname() }})
                {{ post.content }}
            </li>
        {% endfor %}
    </ul>
{% endblock %}

from tipfy import Rule

def get_rules(app):
    rules = [
        Rule('/posts', endpoint='post-listing', handler='apps.blog.handlers.PostListingHandler'),
    ]

    return rules

% /usr/local/google_appengine/dev_appserver.py app

config['tipfy.ext.session'] = {
    'secret_key' : 'just_dev_testH978DAGV9B9sha_W92S',
}

% wget http://bitbucket.org/simplecodes/wtforms/get/tip.tar.bz2
% tar -xvf tip.tar.bz2
% cp -r wtforms/wtforms/ ~/Projetos/gaeseries/app/lib/
% wget http://pypi.python.org/packages/source/t/tipfy.ext.wtforms/tipfy.ext.wtforms-0.6.tar.gz
% tar -xvzf tipfy.ext.wtforms-0.6.tar.gz
% cp -r tipfy.ext.wtforms-0.6/tipfy ~/Projetos/gaeseries/app/distlib

class PostForm(Form):
    csrf_protection = True
    title = fields.TextField('Title', validators=[validators.Required()])
    content = fields.TextAreaField('Content', validators=[validators.Required()])

class NewPostHandler(RequestHandler, AppEngineAuthMixin, AllSessionMixins):
    middleware = [SessionMiddleware]

    @login_required
    def get(self, **kwargs):
        return render_response('new_post.html', form=self.form)

    @login_required
    def post(self, **kwargs):
        if self.form.validate():
            post = Post(
                title = self.form.title.data,
                content = self.form.content.data,
                author = self.auth_session
            )
            post.put()
            return redirect('/posts')
        return self.get(**kwargs)

    @cached_property
    def form(self):
        return PostForm(self.request)

{% extends "base.html" %}

{% block title %}
    New post
{% endblock %}

{% block content %}
    <form action="" method="post" accept-charset="utf-8">
        <p>
            <label for="title">{{ form.title.label }}</label>

            {{ form.title|safe }}

            {% if form.title.errors %}
            <ul class="errors">
                {% for error in form.title.errors %}
                <li>{{ error }}</li>
                {% endfor %}
            </ul>
            {% endif %}
        </p>
        <p>
            <label for="content">{{ form.content.label }}</label>

            {{ form.content|safe }}

            {% if form.content.errors %}
            <ul class="errors">
                {% for error in form.content.errors %}
                <li>{{ error }}</li>
                {% endfor %}
            </ul>
            {% endif %}
        </p>
        <p><input type="submit" value="Save post"/></p>
    </form>
{% endblock %}

% /usr/local/google_appengine/appcfg.py update app

application: gaeseries
version: 3
runtime: python
api_version: 1

handlers:
- url: .*
  script: main.py

% wget http://github.com/mitsuhiko/flask/zipball/0.6
% unzip mitsuhiko-flask-0.6-0-g5cadd9d.zip
% cp -r mitsuhiko-flask-5cadd9d/flask ~/Projetos/blog/gaeseries
% wget http://pypi.python.org/packages/source/W/Werkzeug/Werkzeug-0.6.2.tar.gz
% tar -xvzf Werkzeug-0.6.2.tar.gz
% cp -r Werkzeug-0.6.2/werkzeug ~/Projetos/blog/gaeseries/
% wget http://pypi.python.org/packages/source/J/Jinja2/Jinja2-2.5.tar.gz
% tar -xvzf Jinja2-2.5.tar.gz
% cp -r Jinja2-2.5/jinja2 ~/Projetos/blog/gaeseries/
% wget http://pypi.python.org/packages/source/s/simplejson/simplejson-2.1.1.tar.gz
% tar -xvzf simplejson-2.1.1.tar.gz
% cp -r simplejson-2.1.1/simplejson ~/Projetos/blog/gaeseries/

% mkdir blog
% touch blog/__init__.py

from flask import Flask
import settings

app = Flask('blog')
app.config.from_object('blog.settings')

import views

% touch blog/views.py
% touch blog/settings.py

DEBUG=True
SECRET_KEY='dev_key_h8hfne89vm'
CSRF_ENABLED=True
CSRF_SESSION_LKEY='dev_key_h8asSNJ9s9=+'

from google.appengine.ext import db

class Post(db.Model):
    title = db.StringProperty(required = True)
    content = db.TextProperty(required = True)
    when = db.DateTimeProperty(auto_now_add = True)
    author = db.UserProperty(required = True)

from blog import app
from models import Post
from flask import render_template

@app.route('/posts')
def list_posts():
    posts = Post.all()
    return render_template('list_posts.html', posts=posts)

<html>
    <head>
        <meta http-equiv="Content-type" content="text/html; charset=utf-8"/>
        <title>{% block title %}Blog{% endblock %}</title>
    </head>
    <body>
        {% block content %}{% endblock %}
    </body>
</html>

{% extends "base.html" %}

{% block content %}
<ul>
    {% for post in posts %}
    <li>
        {{ post.title }} (written by {{ post.author.nickname() }})

        {{ post.content }}
    </li>
    {% endfor %}
</ul>
{% endblock %}

from google.appengine.ext.webapp.util import run_wsgi_app
from blog import app

run_wsgi_app(app)

% /usr/local/google_appengine/dev_appserver.py .

from functools import wraps
from google.appengine.api import users
from flask import redirect, request

def login_required(func):
    @wraps(func)
    def decorated_view(*args, **kwargs):
        if not users.get_current_user():
            return redirect(users.create_login_url(request.url))
        return func(*args, **kwargs)
    return decorated_view

% wget http://pypi.python.org/packages/source/W/WTForms/WTForms-0.6.zip
% unzip WTForms-0.6.zip
% cp -r WTForms-0.6/wtforms ~/Projetos/blog/gaeseries/
% wget http://pypi.python.org/packages/source/F/Flask-WTF/Flask-WTF-0.2.3.tar.gz
% tar -xvzf Flask-WTF-0.2.3.tar.gz
% cp -r Flask-WTF-0.2.3/flaskext ~/Projetos/blog/gaeseries/

from flaskext import wtf
from flaskext.wtf import validators

class PostForm(wtf.Form):
    title = wtf.TextField('Title', validators=[validators.Required()])
    content = wtf.TextAreaField('Content', validators=[validators.Required()])

@app.route('/posts/new', methods = ['GET', 'POST'])
@login_required
def new_post():
    form = PostForm()
    if form.validate_on_submit():
        post = Post(title = form.title.data,
                    content = form.content.data,
                    author = users.get_current_user())
        post.put()
        flash('Post saved on database.')
        return redirect(url_for('list_posts'))
    return render_template('new_post.html', form=form)

{% extends "base.html" %}

{% block content %}
    <h1 id="">Write a post</h1>
    <form action="{{ url_for('new_post') }}" method="post" accept-charset="utf-8">
        {{ form.csrf_token }}
        <p>
            <label for="title">{{ form.title.label }}</label>

            {{ form.title|safe }}

            {% if form.title.errors %}
            <ul class="errors">
                {% for error in form.title.errors %}
                <li>{{ error }}</li>
                {% endfor %}
            </ul>
            {% endif %}
        </p>
        <p>
            <label for="content">{{ form.content.label }}</label>

            {{ form.content|safe }}

            {% if form.content.errors %}
            <ul class="errors">
                {% for error in form.content.errors %}
                <li>{{ error }}</li>
                {% endfor %}
            </ul>
            {% endif %}
        </p>
        <p><input type="submit" value="Save post"/></p>
    </form>
{% endblock %}

% /usr/local/google_appengine/appcfg.py update .

application: gaeseries
version: 2
api_version: 1
runtime: python

handlers:

- url: /(?P<a>.+?)/static/(?P<b>.+)
  static_files: applications/\1/static/\2
  upload: applications/(.+?)/static/(.+)
  secure: optional
  expiration: "90d"

- url: /admin-gae/.*
  script: $PYTHON_LIB/google/appengine/ext/admin
  login: admin
 
- url: /_ah/queue/default
  script: gaehandler.py
  login: admin

- url: .*
  script: gaehandler.py  
  secure: optional

skip_files: |
^(.*/)?(
 (app\.yaml)|
 (app\.yml)|
 (index\.yaml)|
 (index\.yml)|
 (#.*#)|
 (.*~)|
 (.*\.py[co])|
 (.*/RCS/.*)|
 (\..*)|
 ((admin|examples|welcome)\.tar)|
 (applications/(admin|examples)/.*)|
 (applications/.*?/databases/.*) |
 (applications/.*?/errors/.*)|
 (applications/.*?/cache/.*)|
 (applications/.*?/sessions/.*)|
 )$

current_user_id = (auth.user and auth.user.id) or 0

db.define_table('posts', db.Field('title'),
                    db.Field('content', 'text'),
                    db.Field('author', db.auth_user, default=current_user_id, writable=False),
                    db.Field('date', 'datetime', default=request.now, writable=False)
                )

db.posts.title.requires = IS_NOT_EMPTY()
db.posts.content.requires = IS_NOT_EMPTY()

def index():
    posts = db().select(db.posts.ALL)
    return response.render('posts/index.html', locals())

{{extend 'layout.html'}}
<h1 id="">Listing all posts</h1>
<dl>
    {{for post in posts:}}
    <dt>{{=post.title}} (written by {{=post.author.first_name}})</dt>
    <dd>{{=post.content}}</dd>
    {{pass}}
</dl>

% /usr/local/google_appengine/dev_appserver.py .

@auth.requires_login()
def new():
    form = SQLFORM(db.posts, fields=['title','content'])
    if form.accepts(request.vars, session):
        response.flash = 'Post saved.'
        redirect(URL('blog', 'posts', 'index'))
    return response.render('posts/new.html', dict(form=form))

{{extend 'layout.html'}}

<h1 id="">
Save a new post</h1>
{{=form}}

% /usr/local/google_appengine/appcfg.py update .

% juju bootstrap
% juju deploy mysql
% juju deploy wordpress
% juju add-relation wordpress mysql
% juju expose wordpress

import subprocess
import traceback

from cloudinit import CloudConfig, util

frequency = CloudConfig.per_instance


def yum_install(packages):
    cmd = ["yum", "--quiet", "--assumeyes", "install"]
    cmd.extend(packages)
    subprocess.check_call(cmd)


def handle(_name, cfg, _cloud, log, args):
    pkglist = util.get_cfg_option_list_or_str(cfg, "packages", [])

    if pkglist:
        try:
            yum_install(pkglist)
        except subprocess.CalledProcessError:
            log.warn("Failed to install yum packages: %s" % pkglist)
            log.debug(traceback.format_exc())
            raise

    return True

#cloud-config
modules:
 - yum_packages
packages: [emacs-nox]

--- cloudinit/util.py 2012-05-22 12:18:21.000000000 -0300
+++ cloudinit/util.py 2012-05-31 12:44:24.000000000 -0300
@@ -227,7 +227,7 @@
         stderr=subprocess.PIPE, stdin=subprocess.PIPE)
     out, err = sp.communicate(input_)
     if sp.returncode is not 0:
-        raise subprocess.CalledProcessError(sp.returncode, args, (out, err))
+        raise subprocess.CalledProcessError(sp.returncode, args)
     return(out, err)

diff -u juju-0.5-bzr531.orig/juju/providers/common/cloudinit.py juju-0.5-bzr531/juju/providers/common/cloudinit.py
--- juju-0.5-bzr531.orig/juju/providers/common/cloudinit.py 2012-05-31 15:42:17.480769486 -0300
+++ juju-0.5-bzr531/juju/providers/common/cloudinit.py 2012-05-31 15:55:13.342884919 -0300
@@ -324,3 +324,32 @@
             "machine-id": self._machine_id,
             "juju-provider-type": self._provider_type,
             "juju-zookeeper-hosts": self._join_zookeeper_hosts()}
+
+
+class CentOSCloudInit(CloudInit):
+
+    def _collect_packages(self):
+        packages = [
+            "bzr", "byobu", "tmux", "python-setuptools", "python-twisted",
+            "python-txaws", "python-zookeeper", "python-devel", "juju"]
+        if self._zookeeper:
+            packages.extend([
+                "zookeeper", "libzookeeper", "libzookeeper-devel"])
+        return packages
+
+    def render(self):
+        """Get content for a cloud-init file with appropriate specifications.
+
+        :rtype: str
+
+        :raises: :exc:`juju.errors.CloudInitError` if there isn't enough
+            information to create a useful cloud-init.
+        """
+        self._validate()
+        return format_cloud_init(
+            self._ssh_keys,
+            packages=self._collect_packages(),
+            repositories=self._collect_repositories(),
+            scripts=self._collect_scripts(),
+            data=self._collect_machine_data(),
+            modules=["ssh", "yum_packages", "runcmd"])

diff -ur juju-0.5-bzr531.orig/juju/providers/common/utils.py juju-0.5-bzr531/juju/providers/common/utils.py
--- juju-0.5-bzr531.orig/juju/providers/common/utils.py 2012-05-31 15:42:17.480769486 -0300
+++ juju-0.5-bzr531/juju/providers/common/utils.py 2012-05-31 15:44:06.605014021 -0300
@@ -85,7 +85,7 @@
 
 
 def format_cloud_init(
-    authorized_keys, packages=(), repositories=None, scripts=None, data=None):
+    authorized_keys, packages=(), repositories=None, scripts=None, data=None, modules=None):
     """Format a user-data cloud-init file.
 
     This will enable package installation, and ssh access, and script
@@ -117,8 +117,8 @@
         structure.
     """
     cloud_config = {
-        "apt-update": True,
-        "apt-upgrade": True,
+        "apt-update": False,
+        "apt-upgrade": False,
         "ssh_authorized_keys": authorized_keys,
         "packages": [],
         "output": {"all": "| tee -a /var/log/cloud-init-output.log"}}
@@ -136,6 +136,11 @@
     if scripts:
         cloud_config["runcmd"] = scripts
 
+    if modules:
+        cloud_config["modules"] = modules
+
     output = safe_dump(cloud_config)
     output = "#cloud-config\n%s" % (output)
     return output

• Juju needs a user called ubuntu to interact with its machines, so you will need to create this user in your CentOS AMI/template.
• You need to host all RPM packages for juju, cloud-init and following dependencies in some yum repository (I haven't submitted them to any public repository):
  • python-txaws
  • python-txzookeeper
  • zookeeper
• With this patched Juju, you will have a pure-centos cloud. It does not enable you to have multiple OSes in the same environment.

• Only one philosopher can hold a fork at a time
• It must be impossible for a deadlock to occur
• It must be impossible for a philosopher to starve waiting for a fork
• It must be possible for more than one philosopher to eat at the same time

package main

import (
    "fmt"
    "sync"
    "time"
)

type Fork struct {
    sync.Mutex
}

type Table struct {
    philosophers chan Philosopher
    forks []*Fork
}

func NewTable(forks int) *Table {
    t := new(Table)
    t.philosophers = make(chan Philosopher, forks - 1)
    t.forks = make([]*Fork, forks)
    for i := 0; i < forks; i++ {
        t.forks[i] = new(Fork)
    }
    return t
}

func (t *Table) PushPhilosopher(p Philosopher) {
    p.table = t
    t.philosophers <- data-blogger-escaped-0="" data-blogger-escaped-1="" data-blogger-escaped-2="" data-blogger-escaped-3="" data-blogger-escaped-4="" data-blogger-escaped-:="range" data-blogger-escaped-_="" data-blogger-escaped-able="" data-blogger-escaped-anscombe="" data-blogger-escaped-artin="" data-blogger-escaped-chan="" data-blogger-escaped-e9="" data-blogger-escaped-eat="" data-blogger-escaped-eating...="" data-blogger-escaped-eter="" data-blogger-escaped-f="" data-blogger-escaped-fed.="" data-blogger-escaped-fed="" data-blogger-escaped-fmt.printf="" data-blogger-escaped-for="" data-blogger-escaped-func="" data-blogger-escaped-getforks="" data-blogger-escaped-go="" data-blogger-escaped-heidegger="" data-blogger-escaped-homas="" data-blogger-escaped-index="" data-blogger-escaped-int="" data-blogger-escaped-is="" data-blogger-escaped-leftfork.lock="" data-blogger-escaped-leftfork.unlock="" data-blogger-escaped-leftfork="" data-blogger-escaped-leibniz="" data-blogger-escaped-len="" data-blogger-escaped-lizabeth="" data-blogger-escaped-lombard="" data-blogger-escaped-main="" data-blogger-escaped-make="" data-blogger-escaped-n="" data-blogger-escaped-nagel="" data-blogger-escaped-name="" data-blogger-escaped-ork="" data-blogger-escaped-ottfried="" data-blogger-escaped-p.eat="" data-blogger-escaped-p.fed="" data-blogger-escaped-p.getforks="" data-blogger-escaped-p.name="" data-blogger-escaped-p.putforks="" data-blogger-escaped-p.table.popphilosopher="" data-blogger-escaped-p.table.pushphilosopher="" data-blogger-escaped-p.table="nil" data-blogger-escaped-p.think="" data-blogger-escaped-p="" data-blogger-escaped-philosopher="" data-blogger-escaped-philosopherindex="" data-blogger-escaped-philosophers="" data-blogger-escaped-popphilosopher="" data-blogger-escaped-pre="" data-blogger-escaped-putforks="" data-blogger-escaped-return="" data-blogger-escaped-rightfork.lock="" data-blogger-escaped-rightfork.unlock="" data-blogger-escaped-rightfork="" data-blogger-escaped-s="" data-blogger-escaped-string="" data-blogger-escaped-struct="" data-blogger-escaped-t.forks="" data-blogger-escaped-t="" data-blogger-escaped-table="" data-blogger-escaped-think="" data-blogger-escaped-thinking...="" data-blogger-escaped-time.sleep="" data-blogger-escaped-type="" data-blogger-escaped-was="">
Any feedback is very welcome.

application: gaeseries
version: 1
runtime: python
api_version: 1

default_expiration: '365d'

handlers:
- url: /remote_api
  script: $PYTHON_LIB/google/appengine/ext/remote_api/handler.py
  login: admin

- url: /_ah/queue/deferred
  script: djangoappengine/deferred/handler.py
  login: admin

- url: /media/admin
  static_dir: django/contrib/admin/media/

- url: /.*
  script: djangoappengine/main/main.py

% python manage.py startapp core

from django.db import models
from django.contrib.auth.models import User

class Post(models.Model):
    title = models.CharField(max_length = 200)
    content = models.TextField()
    date = models.DateTimeField(auto_now_add = True)
    user = models.ForeignKey(User)

% python manage.py createsuperuser

from django.conf.urls.defaults import *

urlpatterns = patterns('',
    ('^$', 'django.views.generic.simple.direct_to_template',
     {'template': 'home.html'}),

    ('^login/$', 'django.contrib.auth.views.login'),
    ('^logout/$', 'django.contrib.auth.views.logout'),
)

{% extends "base.html" %}

{% block content %}

<p>Fill the form below to login in the system ;)</p>

{% if form.errors %}
<p>Your username and password didn't match. Please try again.</p>
{% endif %}

<form method="post" action="{% url django.contrib.auth.views.login %}">{% csrf_token %}
<table>
<tr>
    <td>{{ form.username.label_tag }}</td>
    <td>{{ form.username }}</td>
</tr>
<tr>
    <td>{{ form.password.label_tag }}</td>
    <td>{{ form.password }}</td>
</tr>
</table>

<input type="submit" value="login" />
<input type="hidden" name="next" value="{{ next }}" />
</form>

{% endblock %}

{% extends "base.html" %}

{% block content %}
    Bye :)
{% endblock %}

LOGIN_URL = '/login/'
LOGOUT_URL = '/logout/'
LOGIN_REDIRECT_URL = '/'

class PostForm(forms.ModelForm):
    class Meta:
        model = Post
        exclude = ('user',)

    def save(self, user, commit = True):
        post = super(PostForm, self).save(commit = False)
        post.user = user

        if commit:
            post.save()

        return post

from django.contrib.auth.decorators import login_required
from django.shortcuts import render_to_response
from django.template import RequestContext
from django.http import HttpResponseRedirect
from django.core.urlresolvers import reverse
from forms import PostForm

@login_required
def new_post(request):
    form = PostForm()
    if request.method == 'POST':
        form = PostForm(request.POST)
        if form.is_valid():
            form.save(request.user)
            return HttpResponseRedirect(reverse('core.views.list_posts'))
    return render_to_response('new_post.html',
            locals(), context_instance=RequestContext(request)
    )

def list_posts(request):
    return HttpResponseRedirect('/')

('^posts/new/$', 'core.views.new_post'),
('^posts/$', 'core.views.list_posts'),

{% extends "base.html" %}

{% block content %}
    <form action="{% url core.views.new_post %}" method="post" accept-charset="utf-8">
        {% csrf_token %}
        {{ form.as_p }}
        <p><input type="submit" value="Post!"/></p>
    </form>
{% endblock %}

def list_posts(request):
    posts = Post.objects.all()
    return render_to_response('list_posts.html',
            locals(), context_instance=RequestContext(request)
    )

{% extends "base.html" %}

{% block content %}
<dl>
    {% for post in posts %}
    <dt>{{ post.title }} (written by {{ post.user.username }})</dt>
    <dd>{{ post.content }}</dd>
    {% endfor %}
</dl>
{% endblock %}

% python manage.py deploy

% python manage.py remote createsuperuser

In January I wrote a post for the Rust 2019 call for blogs. The 2020 call is aiming for an RFC and roadmap earlier this time, so here is my 2020 post =]

Last call review: what happened?

An attribute proc-macro like #[wasm_bindgen] but for FFI

This sort of happened... because WebAssembly is growing =]

I was very excited when Interface Types showed up in August, and while it is still very experimental it is moving fast and bringing saner paths for interoperability than raw C FFIs. David Beazley even point this at the end of his PyCon India keynote, talking about how easy is to get information out of a WebAssembly module compared to what had to be done for SWIG.

This doesn't solve the problem where strict C compatibility is required, or for platforms where a WebAssembly runtime is not available, but I think it is a great solution for scientific software (or, at least, for my use cases =]).

"More -sys and Rust-like crates for interoperability with the larger ecosystems" and "More (bioinformatics) tools using Rust!"

I did some of those this year (bbhash-sys and mqf), and also found some great crates to use in my projects. Rust is picking up steam in bioinformatics, being used as the primary choice for high quality software (like varlociraptor, or the many coming from 10X Genomics) but it is still somewhat hard to find more details (I mostly find it on Twitter, and sometime Google Scholar alerts). It would be great to start bringing this info together, which leads to...

"A place to find other scientists?"

Hey, this one happened! Luca Palmieri started a conversation on reddit and the #science-and-ai Discord channel on the Rust community server was born! I think it works pretty well, and Luca also has being doing a great job running workshops and guiding the conversation around rust-ml.

Rust 2021: Interoperability

Rust is amazing because it is very good at bringing many concepts and ideas that seem contradictory at first, but can really shine when synthesized. But can we share this combined wisdom and also improve the situation in other places too? Despite the "Rewrite it in Rust" meme, increased interoperability is something that is already driving a lot of the best aspects of Rust:

• Interoperability with other languages: as I said before, with WebAssembly (and Rust being having the best toolchain for it) there is a clear route to achieve this, but it will not replace all the software that already exist and can benefit from FFI and C compatibility. Bringing together developers from the many language specific binding generators (helix, neon, rustler, PyO3...) and figuring out what's missing from them (or what is the common parts that can be shared) also seems productive.

• Interoperability with new and unexplored domains. I think Rust benefits enormously from not focusing only in one domain, and choosing to prioritize CLI, WebAssembly, Networking and Embedded is a good subset to start tackling problems, but how to guide other domains to also use Rust and come up with new contributors and expose missing pieces of the larger picture?

Another point extremely close to interoperability is training. A great way to interoperate with other languages and domains is having good documentation and material from transitioning into Rust without having to figure everything at once. Rust documentation is already amazing, especially considering the many books published by each working group. But... there is a gap on the transitions, both from understanding the basics of the language and using it, to the progression from beginner to intermediate and expert.

I see good resources for JavaScript and Python developers, but we are still covering a pretty small niche: programmers curious enough to go learn another language, or looking for solutions for problems in their current language.

Can we bring more people into Rust? RustBridge is obviously the reference here, but there is space for much, much more. Using Rust in The Carpentries lessons? Creating RustOpenSci, mirroring the communities of practice of rOpenSci and pyOpenSci?

Comments?

• Thread on Mastodon
• Thread on Twitter

Neste tutorial, será abordado o processo de criação de um dict ou dicionário, a partir de um ou mais dicts em Python.

Como já é de costume da linguagem, isso pode ser feito de várias maneiras diferentes.

Abordagem inicial

Pra começar, vamos supor que temos os seguintes dicionários:

dict_1 = {
    'a': 1,
    'b': 2,
}

dict_2 = {
    'b': 3,
    'c': 4,
}

Como exemplo, vamos criar um novo dicionário chamado new_dict com os valores de dict_1 e dict_2 logo acima. Uma abordagem bem conhecida é utilizar o método update.

new_dict = {}

new_dcit.update(dict_1)
new_dcit.update(dict_2)

Assim, temos que new_dict será:

>> print(new_dict)
{
    'a': 1,
    'b': 3,
    'c': 4,
}

Este método funciona bem, porém temos de chamar o método update para cada dict que desejamos mesclar em new_dict. Não seria interessante se fosse possível passar todos os dicts necessários já na inicialização de new_dict?

Novidades do Python 3

O Python 3 introduziu uma maneira bem interessante de se fazer isso, utilizando os operadores **.

new_dict = {
    **dict_1,
    **dict_2,
}

Assim, de maneira semelhante ao exemplo anterior, temos que new_dict será :

>> print(new_dict['a'])
1
>> print(new_dict['b'])
3
>> print(new_dict['c'])
4

Cópia real de dicts

Ao utilizamos o procedimento de inicialização acima, devemos tomar conseiderar alguns fatores. Apenas os valores do primeiro nível serão realmente duplicados no novo dicionário. Como exemplo, vamos alterar uma chave presente em ambos os dicts e verificar se as mesmas possuem o mesmo valor:

>> dict_1['a'] = 10
>> new_dict['a'] = 11
>> print(dict_1['a'])
10
>> print(new_dict['a'])
11

Porém isso muda quando um dos valores de dict_1 for uma list, outro dict ou algum objeto complexo. Por exemplo:

dict_3 = {
    'a': 1,
    'b': 2,
    'c': {
        'd': 5,
    }
}

e agora, vamos criar um novo dict a partir desse:

new_dict = {
    **dict_3,
}

Como no exemplo anterior, podemos imaginar que foi realizado uma cópia de todos os elementos de dict_3, porém isso não é totalmente verdade. O que realmente aconteceu é que foi feita uma cópia superficial dos valores de dict_3, ou seja, apenas os valores de primeiro nível foram duplicados. Observe o que acontece quando alteramos o valor do dict presente na chave c.

>> new_dict['c']['d'] = 11
>> print(new_dict['c']['d'])
11
>> print(dict_3['c']['d'])
11 
# valor anterior era 5

No caso da chave c, ela contem uma referência para outra estrutura de dados (um dict, no caso). Quando alteramos algum valor de dict_3['c'], isso reflete em todos os dict que foram inicializados com dict_3. Em outras palavras, deve-se ter cuidado ao inicializar um dict a partir de outros dicts quando os mesmos possuírem valores complexos, como list, dict ou outros objetos (os atributos deste objeto não serão duplicados).

De modo a contornar este inconveniente, podemos utilizar o método deepcopy da lib nativa copy. Agora, ao inicializarmos new_dict:

import copy

dict_3 = {
    'a': 1,
    'b': 2,
    'c': {
        'd': 5,
    }
}

new_dict = copy.deepcopy(dict_3)

O método deepcopy realiza uma cópia recursiva de cada elemento de dict_3, resolvendo nosso problema. Veja mais um exemplo:

>> new_dict['c']['d'] = 11
>> print(new_dict['c']['d'])
11
>> print(dict_3['c']['d'])
5 
# valor não foi alterado

Conclusão

Este artigo tenta demonstrar de maneira simples a criação de dicts, utilizando os diversos recursos que a linguagem oferece bem como os prós e contras de cada abordagem.

Referências

Para mais detalhes e outros exemplos, deem uma olhada neste post do forum da Python Brasil aqui.

É isso pessoal. Obrigado por ler!

Eu e o prof. Leonardo Cruz da Faculdade de Ciências Sociais estamos juntos trabalhando no desenvolvimento do Laboratório Amazônico de Estudos Sociotécnicos da UFPA.

Nossa proposta é realizar leituras e debates críticos sobre o tema da sociologia da tecnologia, produzir pesquisas teóricas e empíricas na região amazônica sobre as relações entre tecnologia e sociedade, e trabalhar com tecnologias livres em comunidades próximas a Belém.

No momento estamos com um grupo de estudos montado com cronograma de textos e filmes para trabalharmos e debatermos criticamente. Esse grupo será o embrião para a orientação de alunos de graduação e pós em temas como impacto da inteligência artificial, computação e guerra, cibernética, vigilantismo, capitalismo de plataforma, fake news, pirataria, software livre, e outros.

Aos interessados, nosso cronograma de estudos está disponível nesse link.

E para quem usa Telegram, pode acessar o grupo de discussão aqui.

Quaisquer dúvidas, só entrar em contato!