CSV Challenge

You got your hands on some data that was leaked from a social network and you want to help the poor people.

Luckily you know a government service to automatically block a list of credit cards.

The service is a little old school though and you have to upload a CSV file in the exact format. The upload fails if the CSV file contains invalid data.

The CSV files should have two columns, Name and Credit Card. Also, it must be named after the following pattern:

YYYYMMDD.csv.

The leaked data doesn't have credit card details for every user and you need to pick only the affected users.

The data was published here:

data.json

You don't have much time to act.

What tools would you use to get the data, format it correctly and save it in the CSV file?

---

Do you have a crazy vim configuration that allows you to do all of this inside your editor? Are you a shell power user and write this as a one-liner? How would you solve this in your favorite programming language?

Show your solution in the comments below!

Comments

[Thomas Rayner]

PowerShell to the rescue!

$json = invoke-webrequest 'gist.githubusercontent.com/jorinvo...' | convertfrom-json

$json | select name,creditcard | export-csv "$(get-date -format yyyyMMdd).csv" -NoTypeInformation

[Daniel Coturel]

Excellent, man

[Tobias Salzmann]

ramda-cli:

curl -s https://gist.githubusercontent.com/jorinvo/7f19ce95a9a842956358/raw/e319340c2f6691f9cc8d8cc57ed532b5093e3619/data.json \
| ramda 'filter where name: (complement isNil), creditcard: (complement isNil)' 'map (x) -> x.name + ", " + x.creditcard' -o raw > `date +%Y%m%d.csv`

scala:

import java.io.{BufferedWriter, FileOutputStream, OutputStreamWriter}
import java.text.SimpleDateFormat
import java.util.Date

import io.circe.generic.auto._
import io.circe.parser._

object Data extends App {
  case class CCInfo(name: Option[String], creditcard: Option[String])

  val url = "https://gist.githubusercontent.com/jorinvo/7f19ce95a9a842956358/raw/e319340c2f6691f9cc8d8cc57ed532b5093e3619/data.json"
  val json = scala.io.Source.fromURL(url).mkString

  val infos = decode[List[CCInfo]](json).toOption.get

  val lines = infos.collect{case CCInfo(Some(name), Some(creditcard)) => s"$name, $creditcard"}

  Helper.writeToFile(lines, s"${Helper.formatDate("yyyyMMdd")}.csv")
}

object Helper {
  def writeToFile(lines: TraversableOnce[String], fileName: String): Unit = {
    val writer = new BufferedWriter(new OutputStreamWriter(new FileOutputStream(fileName)))
    for (x <- lines) {
      writer.write(x + "\n")
    }
    writer.close()
  }

  def formatDate(format: String, date: Date = new Date()) = 
    new SimpleDateFormat(format).format(new Date())
} 

[Ayman Nedjmeddine]

A oneliner if you're a linuxer 😉

curl -sSLo- https://gist.githubusercontent.com/jorinvo/7f19ce95a9a842956358/raw/e319340c2f6691f9cc8d8cc57ed532b5093e3619/data.json \
| jq -r '.[] | {name: .name, creditcard: .creditcard} | join(",")' \
> `date +%Y%m%d`.csv

However, there is something you have not mentioned in your post: Should the CSV file have the header line?

If yes, then use this:

echo 'name,creditcard' > `date +%Y%m%d`.csv && \
curl -sSLo- https://gist.githubusercontent.com/jorinvo/7f19ce95a9a842956358/raw/e319340c2f6691f9cc8d8cc57ed532b5093e3619/data.json \
| jq -r '.[] | {name: .name, creditcard: .creditcard} | join(",")' \
>> `date +%Y%m%d`.csv

[Devin Weaver]

This adds quotes.

"Dax Brekke II,1234-2121-1221-1211"
"Brando Stanton Jr.,1228-1221-1221-1431"
"Lacey McDermott PhD,"
"Elza Bauch,"

Maybe adding this sed command:

curl -sSLo- https://gist.githubusercontent.com/jorinvo/7f19ce95a9a842956358/raw/e319340c2f6691f9cc8d8cc57ed532b5093e3619/data.json \
| jq '.[] | {name: .name, creditcard: .creditcard} | join(",")' \
| sed -e 's/^"//' -e 's/"$//' -e 's/\\"/"/g' \
> "$(date +%Y%m%d).csv"

[Richard Metzler]

Doesn't the second solution need a >> in the last line, so the output is appended?

[Ayman Nedjmeddine]

Yes, it does. (Didn't copy the correct version)

Thanks ☺

[Francesco Cogno]

Aaaand Rust :)

Really an overkill for this task but fun nevertheless!

extern crate chrono;
extern crate csv;
extern crate futures;
extern crate hyper;
extern crate hyper_tls;
extern crate serde;
#[macro_use]
extern crate serde_derive;
extern crate serde_json;
extern crate tokio_core;

use futures::prelude::*;
use futures::future::ok;
use tokio_core::reactor::Core;
use hyper::client::Client;
use hyper_tls::HttpsConnector;
use chrono::{DateTime, FixedOffset};
use std::collections::HashMap;
use csv::Writer;
use std::fs::File;


#[derive(Debug, Deserialize, Clone)]
struct Record {
    name: String,
    email: Option<String>,
    city: Option<String>,
    mac: String,
    timestamp: String,
    creditcard: Option<String>,
}

#[derive(Debug, Clone)]
struct RecordParsed {
    record: Record,
    ts: DateTime<FixedOffset>,
}

const FORMAT: &'static str = "%Y%m%d";

fn main() {
    let mut core = Core::new().unwrap();
    let client = Client::configure()
        .connector(HttpsConnector::new(4, &core.handle()).unwrap())
        .build(&core.handle());

    let uri = "https://gist.githubusercontent.com/jorinvo/7f19ce95a9a842956358/raw/e319340c2f6691f9cc8d8cc57ed532b5093e3619/data.json"
        .parse()
        .unwrap();

    let fut = client.get(uri).and_then(move |resp| {
        resp.body().concat2().and_then(move |body| {
            let array: Vec<Record> = serde_json::from_slice(&body as &[u8]).unwrap();
            let mut a_parsed: HashMap<String, Vec<RecordParsed>> = HashMap::new();

            array
                .into_iter()
                .filter(|item| item.creditcard.is_some())
                .map(|item| {
                    let dt =
                        DateTime::parse_from_str(&item.timestamp, "%Y-%m-%d %H:%M:%S %z").unwrap();

                    let rp = RecordParsed {
                        record: item,
                        ts: dt,
                    };

                    let date_only = format!("{}.csv", rp.ts.format(FORMAT).to_string());

                    let ret = match a_parsed.get_mut(&date_only) {
                        Some(ar) => {
                            ar.push(rp);
                            None
                        }
                        None => {
                            let mut ar: Vec<RecordParsed> = Vec::new();
                            ar.push(rp);
                            Some(ar)
                        }
                    };

                    if let Some(ar) = ret {
                        a_parsed.insert(date_only, ar);
                    }
                })
                .collect::<()>();

            a_parsed
                .iter()
                .map(|(key, array)| {
                    println!("generating file == {:?}", key);
                    let file = File::create(key).unwrap();
                    let mut wr = Writer::from_writer(file);

                    array
                        .iter()
                        .map(|record| {
                            let creditcard = match record.record.creditcard {
                                Some(ref c) => c,
                                None => panic!("should have filtered those!"),
                            };
                            wr.write_record(&[&record.record.name, creditcard]).unwrap();
                        })
                        .collect::<()>();
                })
                .collect::<()>();

            ok(())
        })
    });

    core.run(fut).unwrap();
}

[Timur Zurbaev]

PHP:

<?php

$json = json_decode(file_get_contents('https://gist.githubusercontent.com/jorinvo/7f19ce95a9a842956358/raw/e319340c2f6691f9cc8d8cc57ed532b5093e3619/data.json'), true);

$users = array_filter($json, function (array $item) {
    return !empty($item['name']) && !empty($item['creditcard']);
});

$file = fopen(date('Ymd').'.csv', 'w+');

foreach ($users as $user) {
    fputcsv($file, [$user['name'], $user['creditcard']]);
}

fclose($file);

[Michael Orji]

You beat me to the PHP implementation. And your solution is so elegant.

[Devin Weaver]

Since the input JSON could be really large, here is a Node.JS steaming version (using stream-json package):

#!/usr/bin/env node
let fs = require('fs');
let { Transform } = require('stream');
let StreamArray = require("stream-json/utils/StreamArray");
let stream = StreamArray.make();

function escapeCSV(str) {
  if (str == null) { return ''; }
  return /[",]/.test(str) ? `"${str.replace(/"/g, '\\"')}"` : str;
}

class CsvStream extends Transform {
  constructor() {
    super({objectMode: true});
  }
  _transform(chunk, enc, cb) {
    let { name, creditcard } = chunk.value;
    let line = [name, creditcard].map(escapeCSV).join(',');
    this.push(`${line}\n`);
    cb();
  }
}

process.stdin
  .pipe(stream.input);

stream.output
  .pipe(new CsvStream())
  .pipe(process.stdout);

[jorin]

Nice! There is also csv-write-stream then you can save some code :)

[Ryan Palo]

Using the CSV module to avoid any quoting pitfalls. :)

require 'CSV'
require 'date'
require 'JSON'

data = JSON.parse(`curl #{ARGV[0]}`)
filename = Date.today.strftime('%Y%m%d') + '.csv'

CSV.open("#{filename}.csv", 'w') do |csv|
  data
    .select { |item| item['name'] && item['creditcard'] }
    .map { |item| [item['name'], item['creditcard']] }
    .sort
    .each { |item| csv << item }
end

[jorin]

Ruby is still one of the most pretty languages!
Maybe you can use the open(url).read from require 'open-uri' instead of curl to allow it to run on other systems 🙂

Alernatively could look like this:

CSV.open "#{Date.today.strftime '%Y%m%d'}.csv", 'w' do |csv|
  JSON.parse(open(ARGV[0]).read).each { |x| csv << x if x['creditcard'] }
end

[Ryan Palo]

Oh, I like that!

1. I didn't know about those extra options for CSV. Awesome.
2. I didn't know about the open-uri built-in. Also awesome.
3. I love the short and sweet each block! It even feels a little Pythonic, which is nice. Also also awesome!

[Šimon Let]

Oneliner:

curl "https://gist.githubusercontent.com/jorinvo/7f19ce95a9a842956358/raw/e319340c2f6691f9cc8d8cc57ed532b5093e3619/data.json" 2>/dev/null | \
 jq '.[] | .name +","+ .creditcard' --raw-output > `date +"%Y%m%d.csv"`

[Josh Cheek]

A few things to note: cache is a program I wrote that caches command-line invocations, it's to make it cheap to iterate (e.g. so you don't have to hit the network each time) github.com/JoshCheek/dotfiles/blob...

My shell is fish (fishshell.com) which allows multi-line editing, and the parentheses in fish are like backticks in bash, so the > (...) is redirecting the output into a file whose name is the result of the ...

[Florian Rohrer]

Nice post!

import json
from csv import DictWriter

with open("data.json", "r") as f:
    users = json.load(f)

cols = ["name", "creditcard"]
with open("20150425.csv", "w", newline='') as f:
    dw = DictWriter(f, cols)
    dw.writeheader()
    for u in users:
        if u["creditcard"]:
            dw.writerow({k: u[k] for k in cols})

All users share the same date. So I didn't bother and didn't write into separate files.
Another thing, I was going to write "Hey, that's not valid json you are giving us.", because I saw the objects are in a list and that list is not wrapped into an outer object. But my Python parser did not complain, so it turns out valid. You learn something new every day.

[jorin]

Having arrays on the top-level of JSON documents is indeed valid although it is definitely an anti-pattern. By doing so you block yourself from adding any meta information in the future.
If you build an API, you always want to wrap an array in an object. Then you can add additional fields like possible errors or pagination later on.
e.g.

{
  "data": [],
  "status": "not ok",
  "error": { "code": 123, "message": "..." },
  "page": 42
}

[Tobias Salzmann]

Personally, I'd prefer the array in most cases. If I call an endpoint called customers, I would expect it to return an array of customers, not something that contains such an array, might or might not have an error and so on.
If I want to stream the response, I'd also be better off with an array, because whatever streaming library I use probably supports it.

[Tobias Salzmann]

Seems like json can have an array at the root, even according to the first standard: tools.ietf.org/html/rfc4627, section 2