DEV Community: Mišo

Correct exception names in Javascript

Mišo — Wed, 22 Nov 2023 16:56:04 +0000

If you have your own exception hierarchy in the app you probably noticed your perfectly named exceptions show like this in the console or Sentry: Error: something happened and I don't like it. The Error prefix is the name of the exception. I would expect something like this MalformedData: I like different data more. Got: "bad data".

The trick is to set a name attribute of the Error. People do it in every exception in constructor like this.

export class AppException extends Error {}

export class ResourceNotFound extends AppException {
  constructor(
    public resource: string,
    public id: unknown,
  ) {
    super(`The resource '${resource}' with id ${id} is not found.`);
    this.name = 'ResourceNotFound';
  }
}

export class MalformedData extends AppException {
  constructor(data: unknown) {
    super(`I like different data more. Got: ${JSON.stringify(data)}`);
    this.name = 'MalformedData';
  }
}

It is very repetitive and you can forget to do it easily. Also, if you don't need a special constructor for your exception it adds some boiler plate. It can be shortened by setting the name directly:

export class ResourceNotFound extends AppException {
  override readonly name = 'ResourceNotFound';
}

But the ultimate solution for me is to set it once in the top-level exception and let it be derived automatically. Setting it to readonly prevents people overwriting it too.

export class AppException extends Error {
  override readonly name = this.constructor.name;
}

AI tools to try

Mišo — Sun, 30 Apr 2023 17:19:55 +0000

https://learnprompting.org/ - how to write good prompts to get best results from AI chatbots.
https://www.perplexity.ai/ - AI powered search engine.
https://consensus.app/ - AI powered search engine in the research papers.
https://www.d-id.com/ - animated avatars from text to image without any technical knowledge.
https://paintbytext.chat/ - edit images by text.
https://this-person-does-not-exist.com/ - generate random face photo of person.
https://askyourpdf.com/ - upload PDF and ask any questions about its content.
https://lexica.art/ - generate or search previously generated images.
https://play.vercel.ai/ - try and compare the prompt by different models.

Alternative services without tracking

Mišo — Sun, 30 Apr 2023 16:37:52 +0000

https://fonts.bunny.net/ - Google Fonts without tracking by switching the domain only.
https://www.simpleanalytics.com/ - The privacy-first
Google Analytics alternative.
https://nolog.cz/ - Czech movement managing and offering services without collecting user's data.

How I hire Anyor developers to not regret

Mišo — Sat, 09 Apr 2022 18:10:46 +0000

I did quite a lot of technical interviews and I wanted to keep some guidelines to my future myself and improve it over time. The most of the interviews I did was for backend positions (Kotlin, Java, Python, Node.js) but I also did some frontend interviews (TypeScript, React.js).

I believe that no interview gives 100% proof that the candidate is a good fit. The point of my interviews was to know if the candidate has really basic understanding of the technology used and fits into the team personally. But that don't mean to refuse the candidate with different thinking. The diversity is also important in the team. For anything more deep the transition period is needed unfortunately.

I start with some not so interesting preparation notes here but you can jump directly to the technical section.

Preparation

Check the CV profile if the candidate sent one.
Check the GitHub code or example code sent by the candidate. If there is something interesting (positive or negative) make a note so you can talk about it during the interview.
Check LinkedIn, Facebook, Twitter, StackOverflow profiles. I usually google the candidate to get as many info as I can.
Tell the candidate the interview usually takes from 1 to 1.5 hours and he will be writing basic the Python/Java/Kotlin/JS and SQL in an online editor. It's good to set expectations.
Ask the candidate if he can bring his own notebook for it. It's better because he knows the keyboard and settings there so the code writing will be smoother. Writing on the paper or the whiteboard is not good because it is usually not readable and it takes much longer time.
1. If he brings it, then prepare WiFi password for him to connect to the internet during the interview.
2. If the candidate does not bring his own notebook you need to prepare the interview notebook. Always check if it works and does not need updates. Is there EN/DE/... keyboard installed?. Is the battery charged or is there an electric socket in the room? Are WiFi and the browser working? You will need https://codeshare.io/ or http://collabedit.com/ document prepared there (syntax highlighting for the language). You can copy and run the code in https://repl.it/languages. For FE the good tool is CodeSandbox
Prepare anything so you can write notes there about the interview.
The good thing is to set some "secret" signal with other people if the candidate will be really bad and you will want to finish the interview as soon as possible. In that case you simply stop asking more questions, move to the outro and I often just ask him if he has any questions and let him go.
Prepare a glass of clean water for the candidate and ask him if he wants coffee/tea/... :)

Introduction + questions

Interview starts with small talk about the candidate and the company. I usually ask some questions below and tell the candidate about the company/architecture/tech. stack. You don't need to ask all the questions. Just those which make sense of course. And you can ask more questions specific for the candidate.

Introduction of the people in the room and the team.
We have a few questions for you to know you better but feel free to interrupt us and ask your own questions anytime. Or you can start asking if you have something on your mind already.
Did you check something about the company? Maybe our websites? - if yes ask the candidate to tell you what he knows so you can only clarify where he is not completely right. If not just tell him something interesting and not too long :)
What are you working on now?. - If the candidate is a student he can tell you about school and the projects there.
Why are you considering to leave your current company? What do you like/dislike on your current job and/or the project(s)? If the candidate hates the company it's a bad sign. There will always be something what could be better but the candidate should be able to list also good things. The worst of all is when the candidate tells you how all the people there are assholes and he is the only one who saves the whole company every time. It's probably the opposite. I would probably ask for specific example of the problems or the situations when the candidate saved the company. And more than one specific example. Also, it's good to explicitly ask for good things.
What's the most/least exciting thing you worked on? - I usually ask if there were some big issues (least exciting thing :) he caused in the past and what would he change on the issue to make it properly next time?
What do you expect from work here? What makes it an interesting company for you? - You probably want to know how he knows about the company, what he thinks is better here than in other companies, ...
Are you looking for a long term cooperation (years) or the short term (less than a year)? How long do you think you are going to stay here if the job will be OK for you? - You probably want long term candidates. If the candidate tells you he plans to travel around the world in 6 months you should really consider if it's OK for you to spend 6 months with teaching him and then he leaves.
Where do you learn new things about the technology? Blogs, books, twitter, podcasts, RSS, Reddit, StackOverflow, side-projects? - What book are you reading right now or you read the last time? What is your favorite blog/podcast. Simply ask about specifics of the channel he mentions. This is because everybody says they learn and read books. But you want to see which one. Even non-technical. Maybe even if you don't hire him at the end he can recommend you something inetresting :)
Is there anything what would annoy you in the office? Dogs, open-space, air conditioning, noise, ... Is there anything you need to have in the office?
Was there any disagreement about what technology should be used, or about the process in the company you've been involved? How was it resolved? - This is something I don't ask often, but if I feel the candidate is more conflicting type or he mentions something from the past I want to know something about these things. He should not be stubborn, but assertiveness is good.

Introduce the company tech. stack

After all the questions it's good to tell the candidate again something about the company and more in-depth technologies used.

The languages and the main frameworks used (Kotlin, Python, Spring, Flask, Celery, React.js, Redux, TypeScript, ...). It's good to mention also versions if that matters.
GitHub/GitLab/BitBucket/Space
CI - GitHub, GitLab, Travis, Jenkins, ....
Slack, M$ Teams, Space, ...
Trello, Jira, Product board, Redmine, Asana, ...
ELK, Graylog, Grafana
Intellij IDEs/VS Code

Ask if the candidate knows the tools. If no, it does not matter, but if yes, ask about how he used it in the work. The big warning sign is if the candidate knows everything. It's so called superman (with small s) and you should ask him about really specific things like:

In what project did you used the technology? What was the use-case?
Why did you choose this technology and not any other similar?
What were the problems you had with the technology?
What do you like about the technology?
When would you use it again and when not?

If he knows the answers on all of them then it's probably Superman (with capital S :) but really make sure. If you hire a superman it's really painful to fire him because they know everything in theory but nothing in practice. Of course every new person should be hired only if you are at least 90% sure you want to work with the candidate, but the candidate who knows every technology is very suspicious.

Hard core skills

This is a part where I really want to see what the candidate knows and how confident he is with his decisions and knowledge. Sometimes (OK, almost every time) I play dumb here and pretend I don't know about the answer the candidate is asking. Instead, suggest to the candidate to Google it - it's part of our job anyway. If the candidate is really confident and I would like to stress him a little I would ask him if he can find the bug in the code even if the code is absolutely OK. I ask questions like: "Are you sure? Any flaw?" To see how he reacts. But I tell him at the end that the code is really OK and it is nice job :)

If you spot any minor issue, like typo in the variable name, just tell the candidate. Don't stress him/her with silly things that IDE can warn him about. This is not a school. But don't answer him a questions like "Is this OK? Is it like that?" If (s)he is not sure about what he wrote he can Google it. When the candidate is writing the code you see how it goes immediately. My strategy is to make the requirements harder and harder until I hit the spot when the candidate does not know. This spot is different for everybody. Simply stop asking when you feel you know where is the upper limit of the candidate's skills. Bellow are the questions I ask along with my favorite exercise tasks.

Python/Java/Kotlin/TypeScript

Java/Kotlin: What is the (functional) interface and why does it exist? What's difference between interface and abstract class? - If the candidate does not what the interface is, I would not waste the time and don't hire the candidate. The good place for the secret signal and finish the interview as soon as possible. Otherwise keep going to check the candidate's seniority :)
What is the decorator (Python/JS) / annotation (JVM)? Can you name any? Did you write your own already? - This question is about to find out if the candidate knows some more advanced features of the language, but not too advanced. I can decide within one minute based on answers if I can give him harder coding exercise or not. Also if it makes sense to ask about generics during the coding.
- Java/Kotlin - @Override, @SuppressWarnings, @Deprecated in Kotlin, @JsonProperty from Jackson
- Python - @property, @classmethod, @app.route from Flask
- Javascript - @connect from Redux, @Component from Angular, @withRouter from React router
When would you use thread and when the process to scale the application and why?
(Python) Do you know GIL? Can you tell us about it?
Do you know any library, standard or 3rd-party, to scale your application?
Do you have experience with tests? Which ones (unit, integration, E2E)? What is your favorite framework?

Code

Write function find_max({1, 2, 3, 4, 5}) (or find_min) that accepts an array of numbers and returns maximum (or minimum) from them.
Can you find any issues in your implementation? - Common bugs here are:
1. People don't consider empty array can come - results in the Out-of-range errors.
2. Sometimes the candidate does not know what to do when the array is empty - raising an exception or returning null is OK IMHO. NaN is kind of half-OK because it does not work with other objects (read later). I try to ask for something else. But returning some special string or INT_MAX/INT_MIN/0 or another weird things are no-go. Any number can be the valid result so we can't use it like special value for empty array.
3. People don't consider negative/real numbers - these are also numbers and the function should work with them.
4. People try to find some MIN_INT/MIN_FLOAT or something to initialize the maximum value - the 1st item from the array should picked instead because it can be Decimal number or complex number or we could change the function to accept also DateTime and there is no minimal/maximal value.
Can you write some basic tests for your code? What values makes sense to test? How many tests are enough?
If the basic code is OK and has no issues I change requirements a little. For example, what if I wanted to call the function like find_max(1, 2, 3, 4, 5) what do I need to change? You can use Google if don't know. The right term to search is "variable arguments".
- Java private static int findMax(int... items)
- Kotlin fun findMax(vararg items: Int): Int
- Python def find_max(*items)
- TypeScript function findMax(...items: number[])
What if I would like to use the function for non-number objects? Can I do it? How? I want to hear about generics, comparators in Java or magic methods (for operator overloading) in Python/Kotlin/TypeScript.
- Java private static <T> T findMax(T... items, Comparator<T> comparator) and then if (comparator.compare(item, maximum) > 0) { maximum = item; }
- Kotlin fun <T> findMax(items: Array<T>, isFirstGreater: (a: T, b: T) -> Int): T and then if (isFirstGreater(item, maximum)) { maximum = item }
- Python def find_max(items, is_first_greater) and then if is_first_greater(item, maximum): maximum = item

/** Kotlin */
fun main(args: Array<String>) {
    assert(findMax(arrayOf(1, 2, 3, 4, 5), { a: Int, b: Int -> a > b }) == 5)
}

fun <T> findMax(items: Array<T>, isFirstGreater: (a: T, b: T) -> Boolean): T {
    if (items.isEmpty()) {
        throw IllegalArgumentException("Empty array is not allowed")
    }

    var maximum = items[0]
    for (item in items) {
        if (isFirstGreater(item, maximum)) {
            maximum = item
        }
    }

    return maximum
}

class Main {
  /** Java */
  public static void main(String[] args) {
    Integer[] numbers = {1, 2, 3, 4, 5};
    assert(findMax(numbers, (a, b) -> a - b) == 5);
  }

  private static <T> T findMax(T[] items, java.util.Comparator<T> comparator) {
    if (items.length == 0) {
        throw new IllegalArgumentException("Empty array is not allowed");
    }

    var maximum = items[0];
    for (final var item : items) {
        if (comparator.compare(item, maximum) > 0) {
            maximum = item;
        }
    }

    return maximum;
  }
}

import pytest

def find_max(*items, is_first_greater=lambda a, b: a > b):
    if not items:
        raise ValueError("Empty array is not allowed")

    maximum, *items = items
    for item in items:
        if is_first_greater(item, maximum):
            maximum = item

    return maximum

assert find_max(1) == 1
assert find_max(1, 2, 3, 4, 5) == 5
assert find_max(-1, -2, -3, -4, -5) == -1
assert find_max(42, 42, 42, 42, 42) == 42
with pytest.raises(ValueError):
    find_max()

If I feel the developer is more experienced or he claim he is senior in CV, or has "excellent" Python/Kotlin skills I have more advanced task prepared - length of the longest continuous non-descending subsequence (sub-array). I ask him if he knows what it is, if not he can google it or you can tell him. The steps are the same as for the previous example. I can add generics or comparator. Also the common bugs are usually similar. But people also forget about previous = item often :)

def longest_increasing_subsequence(items, is_first_increasing=lambda a, b: a >= b):
    if not items:
        return 0  # longest sequence for empty list is zero

    length = 1
    max_length = 1
    previous, *items = items
    # for previous, item in zip(items, items[1:]):
    for item in items:
        if is_first_increasing(item, previous):
            length += 1
        else:
            max_length = max(max_length, length)  # can be if length > max_length: max_length = length
            length = 1

        previous = item

    return max(max_length, length)

assert longest_increasing_subsequence([]) == 0
assert longest_increasing_subsequence([0]) == 1
assert longest_increasing_subsequence([1, 2, 3, 4, 5]) == 5
assert longest_increasing_subsequence([-1, -2, -3, -4, -5]) == 1
assert longest_increasing_subsequence([-5, -4, -3, -2, -1]) == 5
assert longest_increasing_subsequence([42, 42, 42, 42]) == 4
assert longest_increasing_subsequence([1, 2, 1, 0]) == 2
assert longest_increasing_subsequence([1, 2, 1, 2, 3]) == 3

Can the candidate design some relational DB? Postgres, MySQL? The candidate doesn't need to know the DDL for the tables from memory. Can just write what are the tables, columns and types - pseudo-code. You have system with users and departments. User has name and the salary and department has only name. User can be in no department or in more than one.

CREATE TABLE "user" (
    id BIGINT PRIMARY KEY,
    name VARCHAR NOT NULL,
    salary NUMERIC NOT NULL
);

CREATE TABLE "department" (
    id INT PRIMARY KEY,
    name VARCHAR NOT NULL
);

CREATE TABLE "user_in_department" (
    id_user BIGINT REFERENCES "user" ("id"),
    id_department INT REFERENCES "department" ("id")
);

Can you find all users with salary bigger than 100k? SELECT * FROM "user" WHERE "salary" > 100000
Can you find all users with name starting with "M"? SELECT * FROM "user" WHERE "name" LIKE 'M%' I meant any "m" ;) ILIKE or "name" LIKE 'M%' OR "name LIKE 'm%'. There is also function to search substring in PG position or the candidate can use regular expressions.
Can you find all names of the user who are not in any department? SELECT "user"."name" FROM "user" LEFT JOIN "user_in_department" ON "id_user" = "id" WHERE "id_user" IS NULL
Can you find users who are in "IT" department? SELECT "user"."name" FROM "user" JOIN "user_in_department" ON "id_user" = "user"."id" JOIN "department" ON "id_department" = "department"."id" WHERE "department"."name" = 'IT'
Can you list the average salary in every department? Including average salary of the people without assigned department? SELECT "deaprtment"."name", AVG("salary") FROM "user" LEFT JOIN "user_in_department" ON "id_user" = "user"."id" LEFT JOIN "department" ON "id_department" = "department"."id" GROUP BY "department"."name"
Can you sort departments by the count of the users? The biggest departments first? If there is the same amount of the users can then alphabetically according to department name? SELECT "deaprtment"."name" FROM "user" JOIN "user_in_department" ON "id_user" = "user"."id" JOIN "department" ON "id_department" = "department"."id" GROUP BY "department"."id" ORDER BY COUNT("user"."id") DESC, "department"."name"

JavaScript (frontend)

What kind of object inheritance is used in JS and how it works?
What is "pure" function? Why we should use it?
What is "lambda" function? How is it written? What are (dis)advantages?

Code

Write function sum(a, b) that returns a + b if 2 arguments are given and another binded function if only one is given so I can call it either as sum(2, 3) === 5 or as sum(2)(3) === 5. What if we would like to support impossible number of parameters like sum(2, 3, 4)(4, 5)(1, 1, 1, 1)(3, 7)() === 32.
What happens when you execute the following code? How would you fix the code if we would like to alert all values in i?

for (var i = 0; i < 3; i++) {
    setTimeout(function () { alert(i) }, i * 1000)
}

Find even minimum in the array of the numbers. Can you write it in the functional way? (if the candidate does not know how, you can suggest him to use Array.filter, Array.map, Array.reduce).
If the candidate is really good and confident, try to ask what will be stored in numbers in this code const numbers = ["10", "10", "10"].map(parseInt)? The correct answer is [10, NaN, 2] because parseInt(number, radix) accepts also radix from array index so the call will be ["10", "10", "10"].map((value, index) => parseInt(value, index) and radix = 1 does not exist and 10 in radix = 2 is 2.

Outro (practicalities)

This is the section I skip if I know right away we don't want the candidate because we all gave each other the secret signal.

When can you start working with us?
What contract do you prefer?
What is your expected salary? In gross (or net if he's student - students don't pay taxes if they meet certain conditions :).
List the company benefits. It's questionable if you want to list the things everyone in the industry has like flexi-time, home-office. It's good to mention some activities people in the company do like football, board games, movie nights.

More sources:

Python chained comparisons with == and !=

Mišo — Tue, 26 Oct 2021 08:21:23 +0000

IDEA (PyCharm) suggested me today to simplify one condition in Python count > 0 and count != prev_count. I was asking myself how? And it was transformed to 0 < count != prev_count.

I knew this is possible in Python but because every example out there always shows something like min < value <= max I somehow never realized this is possible too. It makes sense because according to documentation all comparison operators are supported for chained comparison. Have it in mind next time :)

Online interactive playground for Nginx and more

Mišo — Wed, 06 Oct 2021 19:20:00 +0000

A few days ago I read an article about the ability to test Nginx configs online. In your browser! I use Caddy these days but there are still some "old" Nginx servers running so I need to change them from time to time. While Caddy has a super easy setup that's not true for Nginx. Online Nginx config generator from Digital Ocean is a great help but to test the config and experiment while tuning it for a specific application is even better. But that's not all. There are more similar playgrounds in the article, so I decided to pin them here for my future self. Feel free to suggest some if you know more.

Codeapi as interactive version of Learn X in Y minutes
nginx playground and nginx location matcher
SQL playground and DB fiddle or SQLime
postgres.new or PostgreSQL in the browser
Regex101 or Regexr
HTTPbin, CORS playground and bad SSL

The living list of "tech" I would like to try

Mišo — Sun, 08 Nov 2020 13:40:26 +0000

Once in a while, I read about some new technology that excites me. But I can't just jump in and if I don't make a note then I can't remind the name when it would be handy. So I decided to gather it here and when I start doing something where I can use the technology I can come here to just click and decide if it is a good fit for my current needs.

GraalVM - Fast (not only) JVM runtime written in Java. Runs and compiles Java, Kotlin, Ruby, C/C++ (or any LLVM based language), Python, JavaScript, ... in one VM and even allows to use all of these languages in one codebase without the runtime overhead. In fact, it is faster than native C bindings from Node.js, Ruby, ... it seems like a great future to me. I could develop my React app in Kotlin and use some ML Python library here and some secure and fast Rust library to hash passwords there while targeting browsers and using critical parts in WebAssembly.

Rust - very interesting alternative to C++ and Go. Very fast, secure by default, and with a mix of the best of both FP and OOP worlds.

Zig - next gen C language that claims to be better compiler for C than C compilers. Consider comparison with https://vlang.io/

Deno with https://alephjs.org/ as an alternative to Next.js.

Elm or fp-ts/io-ts - I was cought by Elm's no runtime exceptions, but TypeScript feels more natural these days.

FaunaDB - NoSQL relational ACID database with native GraphQL API.

Firebase - Set of services (auth, store, hosting, ...) built on top of Google cloud.

Supabase - an open-source alternative to Firebase built on top of PostgreSQL. And maybe also other databases in the future.

Neon - Truly serverless PostgreSQL with generous free tier.

YugabyteDB or CockroachDB, TerminusDB, or HarperDB - simply some alternative to PostgreSQL ready for a really big data. I have quite an experience with ElasticSearch and very little MongoDB, but those do not feel right. Especially the lack of JOINs and no ACID guarantee.

evolu.dev - Privacy-focused local‑first platform that scales. Basically local-first E2EE SQLite DB compiled to WebAssembly with reactive state. Synced to other devices through server. Uses https://kysely.dev/ to query data in a type-safe way.

Dolt - git for relational data. Versioned database with branching all of the good from git.

Airtable - part spreadsheet, part database.

Postgres.ai's Joe bot - Joe bot, an SQL query optimization assistant. Joe works on top of the Database Lab. Every time when an engineer starts communicating with Joe, a new full-size copy of the database is provisioned. This process is fully automated and takes only a few seconds, even for multi-terabyte databases. Such database copies are called “thin clones” because multiple clones share the same data blocks, so provisioning is super fast, and disk space consumption is very low. The clones are fully independent, so developers can modify databases. Finally, SQL execution plans are identical to production, which makes it possible to troubleshoot and optimize queries reliably without involving production databases. Thin provisioning in seconds thanks to copy-on-write (CoW) provided by ZFS and special methodology of preparing PostgreSQL database snapshots. There is also an option to use LVM instead of ZFS.

Steampipe or AnyQuery is Foreign Data Wrapper for PostgreSQL that automatically unifies APIs from various sources and exposes them as a set of tables/views. You can fetch data from API through your DB. Something similar is Superface that is another attempt to unify APIs into one cohesive library.

React Testing Library - the only right way to test React.js apps without implementation details.

XState and https://css-tricks.com/model-based-testing-in-react-with-state-machines/ - alternative to Redux with the ability to draw your business logic as a chart.

Redux-toolkit - Redux with boilerplate reduced.

Snowpack or Vite - Fast replacement for Webpack used for development. Production build still uses Webpack, but development is blazingly fast.

bundle-wizard - webpack analyzer for live websites

Caddy server - HTTPS first, a no-boilerplate alternative to Nginx written in Go. Actually, I use this one already and I am pretty happy with it. The auto TLS certificate renewal is awesome.

Glitch - have no idea what it is but I hit it regularly and it seems like some hybrid of Vercel/Netlify and CodeSandbox/repl.it.

HTMX - minimalistic HTML framework for enhancing static pages with interactivity. Similar to Signals and snippets in Nette.

Gatsby - Next.js alternative for building static sites. But I want to just explore it because I think Next.js is still better. But Gatsby is known for its amazing plugins.

Blitz.js - full-stack React framework. Good old web framework without REST/GraphQL API but with the power of React.js.

Fly.io - cloud without the ops. Not like AWS, GCP, Azure and others.

LogRocket - FullStory and Sentry in one package. There are self-hoste alternatives, even open source, https://www.rrweb.io/ and https://openreplay.com/.

PostHog - Mixpanel, Sentry, AB testing, full story, and others all together in single package. With a generous free tier.

bearer.sh - third-party APIs real-time monitoring in a few lines of code.

StackBlitz - full stack Node.js dev environment running in the browser. Node.js is compiled to WebAssembly to run in the browser. Alternatives are Codespaces and GitPod but I am excited about StackBlitz because it runs in the browser, even backend, has virtual TCP stack, .... On the other hand it supports only Node.js now, and not Kotlin, Python, ... as GitPod.

Space - kind of GitHub, Slack, Wiki, ... combined into one application. The presentation looks really great and it's from JetBrains so it could really cool app.

Super productivity - time tracker and TODO list on steroid. I use Toggl currently, but I would like to have something with integration to GitHub and Trello.

https://gitprime.com - now bought by https://www.pluralsight.com/product/flow - advanced developer metrics

Kialo - discussion with thread supporting arguments and discovering fallacies. Discussion as a tree of pros/cons.

https://pipedream.com/ - like Zappier/Make but with full code control. https://www.napkin.io and https://www.val.town are other alternatives.

Huginn - self-hosted Zappier/IFTTT. There is also one alternative in n8n.

ZeroMQ - broker less messaging framework or sockets on steroids.

Blinkist - all the best book summaries in 15 minutes. It's a little bit controversial because it's questionable if you can get all from the book in a summary from a stranger, but worth to try because I think some books are really good but too verbose for me.

Blockchain - just to work more closely with the crypto stuff.

Full-text search of all git branches

Mišo — Tue, 06 Oct 2020 13:49:25 +0000

When I want to quickly search for something in the code I use GitHub search page. With the prefixes like org:<organization-name> it's really great and I have even defined custom browser search engines for different organizations. Then I can simply write something like org "127.0.0.1" into my browser and it finds all localhosts in my org in all repositories.

The problem is that GitHub indexes only the main branch (whatever you name it). But when there are more branches with active development I want to search even those. And that's why I have a simple batch script to achieve that. Now I can go to any local repo and run git-search "127.0.0.1" or git-search "\b_source\b" and I get all unique lines matching the RE pattern. Here is the magic script to do that.

@echo off
rem https://stackoverflow.com/questions/7151311/using-git-how-could-i-search-for-a-string-across-all-branches
git fetch origin
git remote prune origin
git branch --remote | grep -v " -> " | xargs git grep %* | cut --delimiter=":" --fields=2,3 | uniq | grep --color=always -E %*

Is an exception less (pure) than a Result?

Mišo — Thu, 17 Sep 2020 18:07:46 +0000

Recently I saw a video about the Railway Oriented Programming and I can't help myself but I ask "why"? So I searched and found Against Railway-Oriented Programming from the same author. Still, I have to ask: "Why should I teach how to compose functions that don't match when I can simply throw an exception and have all of this for free? And it is not just this video. With recent OOP hate and functional passion, I hear it all the time how the explicit result is better because then the function is pure and returns always the same output data for the same input data. And then I come to the question. Is that different from the exceptions?

From what I know the function is pure when it returns the same result for the same arguments. And if the function always throws the same exception for the same input arguments it is pure. Even reducers in Redux are considered pure if they don't have side-effects and they throw exceptions all the time (famous undefined is not an object/function/...). So there has to be something else. Something why one would like to abandon the simplicity of throwing exceptions with the whole railway and even call it a better approach. Below is an example to try to find it.

type Result = 
  | {success: true, processedData: ProcessedData}
  | {success: false, error: string};

/**
 * Pure function with the "Result" type.
 */
function pureFunction(user?: User, data: Data): Result {
  if (!user) {
    return {success: false, error: 'User is not authenticated'}
  }

  return {success: true, processedData: {...data, processed: true}}
}

/**
 * The same function but "returning" an error as an exception.
 */
function maybePureFunction(user?: User, data: Data): ProcessedData {
  if (!user) {
    return throw new UserNotAuthenticated(user)
  }

  return {...data, processed: true}
}

Can you spot the difference? I can. It's hidden in types. I don't know any language out there checking the types of the exceptions. Java somehow tried with the checked exceptions and failed, Python don't plan it. Nobody wants to write all the possible exceptions thrown from every function (sorry, method). And still, it is the same as to write all the return types. So this is a big advantage of the Return type - type safety. The compiler will tell you that you don't handle some possible error but don't warn you about uncaught exceptions.

And then I had to ask again. Why? Why can't we have both advantages? Why can't I just throw an exception and let the compiler infer the exceptions for the function and warn me about possibilities? Then I don't need to learn how to combine functions by bind, >>=, etc. I think the idea of checked/unchecked exceptions in Java is good. Only I would name them BusinessException and ProgrammerError or something like that. And they would be inferred by the compiler.

In fact, I am not afraid to consider exceptions to be algebraic data types, sum types or enums as defined by Rust. Here is my ideal API.

try {
    return maybePureFunction(user, data)
} catch (e) {
    // here the compiler knows "e" is "UserNotAuthenticated"
    return DATA_FOR_ANONYMOUS_USER
}

// OR

// compiler infers here the function return type is kind-of "Result<ResourceNotFound | DangerousError, string>"
function something(fileName: string): string {
    try {
        return dangerousCall(fileName)  // throws "DangerousError" unhandled here
    } catch (e: FileDoesNotExist | InsufficientPermissions) {
        throw new ResourceNotFound(fileName);
    } catch (e: FileLocked) {  // isn't this actually pattern matching?
        return ""
    }
}

What do you think about it? And more importantly, do you know such a language where exceptions are a proper part of the type system and are treated as a result of the function?

Interesting sources

Python features by version

Mišo — Mon, 24 Aug 2020 19:10:14 +0000

From time to time I need to poke someone to use a new version of Python but I couldn't find the list of main features one gets by this. There is https://docs.python.org/3/whatsnew/index.html but that is too much info. So here is my concise list. Also, there is an EOL (End of Life) date for every version in the brackets. More visual EOF site at https://devguide.python.org/versions/

Python 3.13 (2029-10)

New colorful interactive REPL
Experimental GIL removal
Experimental JIT compiler
Support for mobile platforms like iOS and Android soon

Python 3.12 (2028-10)

New Type Parameter Syntax for generics in functions def max[T](args: Iterable[T]) -> T: classes class list[T]: and types type Point[T] = tuple[T, T]
Nested f-strings f"{f"{f"{f"{f"{f"{42-36}"}"}"}"}"}"
A per-interpreter GIL

Python 3.11 (2027-10)

between 10-60% speedup
Fine-grained error locations in tracebacks
Enhanced error locations in tracebacks
Support for parsing TOML files in the Standard Library

Python 3.10 (2026-10)

Pattern matching

    match point:
        case Point(x=0, y=0):
            print("Origin is the point's location.")
        case Point(x=0, y=y):
            print(f"Y={y} and the point is on the y-axis.")
        case Point(x=x, y=0):
            print(f"X={x} and the point is on the x-axis.")
        case Point():
            print("The point is located somewhere else on the plane.")
        case _:
            print("Not a point")

Type alias StrCache: typing.TypeAlias = "Cache[str]"
Union type operator def square(number: int | float) -> int | float:
Better error messages with precise line numbers for debugging and other tools
Parenthesized context managers

with (
    CtxManager1(),
    CtxManager2()
):
    ...

Python 3.9 (2025-10)

Dictionary merge assert {"a": 1} | {"b": 2} == {"a": 1, "b": 2}
New string methodsstr.removeprefix(prefix) and str.removesuffix(suffix)
Builtin Generics list[dict[str, int]] vs. typing.List[typing.Dict[str, int]]
New PEG parser
New module zoneinfo for timezones information

Python 3.8 (2024-10)

The walrus operator if (n := len(a)) > 10: print(f"List is too long ({n} elements, expected <= 10)")
Positional-only parameters len(obj, /) which prevents awkward calls such as len(obj="Hello")
Self documenting f-strings f"{user=} {member_since=}" # "user='eric_idle' member_since=datetime.date(1975, 7, 31)"
cached_property is finally in the standard library

Python 3.7 († 2023-06)

@dataclass
class Point:
    x: float
    y: float
    z: float = 0.0

p = Point(1.5, 2.5)
print(p)   # produces "Point(x=1.5, y=2.5, z=0.0)"

Built-in breakpoint function

Python 3.6 († 2021-12)

f-strings f'Hello {name}'
undescores in number literals 10_000_000
variable annotations variable: Type = 29
asynchronous generators and comprehensions

async def ticker(delay, to):
    """Yield numbers from 0 to *to* every *delay* seconds."""
    for i in range(to):
        yield i
        await asyncio.sleep(delay)

result = [i async for i in aiter() if i % 2]
result = [await fun() for fun in funcs if await condition()]

New module "secrets"
Type dict is more compact and ordered by default

Python 3.5 († 2020-09-30)

async/await syntax
matrix multiplication operator M @ N
Additional Unpacking Generalizations print(*[1], *[2], 3, *[4, 5]) and function(**{'a': 1, 'c': 3}, **{'b': 2, 'd': 4})
New enum for HTTP statuses http.HTTPStatus.OK
Type hints typing.Dict[str, typing.Any]

Python 3.4 († 2019-03-18)

Python 3.3 († 2017-09-29)

Built-in virtualenv
Generator delegation yield from another_generator()
Exception context suppresion raise Exception("Ooops") from None
Explicit unicode literals are back to make migration from Python 2 easier u"I am unicode in both Python 2 and 3"
New module unittest.mock
New module ipaddress

Python 3.2 († 2016-02-20)

¯_(ツ)_/¯

Python 3.1 († 2012-04-09)

OrderedDict
with handles multiple context managers in single statement

Python 3.0 († 2009-06-27)

Big pile of incompatibility

My language score-board

Mišo — Mon, 13 Jul 2020 16:29:39 +0000

From time to time someone asks me about my opinion about the programing languages I "know". Here is my ordered list as I would recommend them to learn or choose to build something. And to clear the things a little: I am a full-stack web developer. You can look at this as my personal TIOBE index.

Kotlin

When I read/write in it I have a feeling like I could use Python syntax with extra features on the JVM platform. And that is great symbiosis. Support for WebAssembly and the compilation to JavaScript with the React.js in the Kotlin DSL makes this language my replacement candidate for Python & TypeScript in the future. And Google agrees with me 😉

Python

The most readable syntax, plenty of great libraries, and a good community. I use it professionally and also personally for the automation (with the help of the Fabric). It is considered slow but the critical libraries are written in C so it does not matter much IMHO.

Go

Actually, I never did anything in this language except for the The Go Playground. But I like static "duck typing", defer, go-routines. I guess the lack of the exceptions was surprising for me, but the missing generics is probably the reason I never really tried it for a bigger project.

Typescript

This is my choice when I need to do any app in the browser. I think it's the best, typed, browser language today. Even if I can imagine a much better type-system I really like how it can be incorporated incrementally and use familiar JS syntax. I don't want to go back to pure JS. But still, I am ready to switch it for something better in the future because it's far from the best.

Java

I worked in the ecosystem for a few years. Java definitely gave me a range of visions I use in Python and other languages. But the language itself is simply old. The new features just copy other languages that are years ahead. The problem is that at the same time the legacy is not let in the past. The thing I didn't like the most was the community. I felt like there are a lot of arrogant people who believe Java is the best forever and if you can't solve the problem without the help you are a bad programmer. Also, I had that feeling that a lot of things are overengineered (not ready for the common simple cases). I would suggest learning Kotlin.

PHP

The language I started with seriously (I don't count Baltík, Assembler, and Pascal in the school). In the time, it was kind of starter language the JS is today, but even if I don't work with it a few years now I see it evolved into a good language with a nice community. I have learned a lot from Nette comunity. Still, I would suggest to learn Python instead.

Javascript

The starter language of today. I use it just because it is the only widely supported language in the browser. The strength is that it is also in the server so you can share the code. The built-in async support is great and fast for the IO-bound apps and the path from when I started hacking with it in 2005 and now is incredible. A lot of great tools and progress were made and even language evolved into something totally different.

Elm

Elm is many things better than JS/TS. Maybe in all. But there is a missing ecosystem of the libraries the JS has. I recommend to watch at least the community and some videos about the nicest error messages, NO runtime errors, type system, and other ideas. I learned a lot from it and I am not alone. A lot of good things in JS is inspired here.

Haskell

The real functional programming language. And I don't mean functions + .map/.filter/.reduce as in JS. The built-in immutability, great type-inference, pattern-matching, recursion, lazy evaluation, pure functions, Monads, ... I tried the language once at the school for one semester and wrote one project. And even I would not recommend it for the beginner it was really great to learn it. It opened my eyes in many ways and I try to follow some patterns from it until today in other languages.

C++

Another dinosaur as Java. At the school, I was glad every time I could use it but it was just because the alternative was C. It's widely used and I don't believe it will disappear anytime soon, but I guess Rust or Go will be growing and cutting the C++ usage.

C

Simple language with good access to the bare metal. That's why it is not used too much today when the web dominates. If you don't need any specific feature or are not forced to use it use Rust/Go/C++ instead.

Prolog

The interesting language that is able to solve many goal-oriented problems with the strict rules very efficiently. Think of any logic puzzle for example. But I tried to solve the maze game and was very frustrated because it was hard to debug. Not my cup of cake 😀

VHDL

Just used it at the school for one project and it was not my taste of cake :) This is something I don't know anything about. Just here because I want to remember I used it.

Assembler

Hard to learn, hard to use, but with the good macros, it feels like C with more control. But still, I would probably use Rust/Go/C++/C with some bits of ASM today instead.

Perl

Write-only language. Really unreadable mess. Try to avoid it. Fortunately, it is slowly dying.

Brainfuck/chicken/whitespace

Not real languages, but crazy experiments. I hope nobody is forced to do real work with these.

Conclusion

If you are backend/mobile/desktop developer try Kotlin and/or Python. If you are full stack and are brave enough try Kotlin even on frontent. Still brave but not so much? Try Elm. Otherwise, Typescript/React.js combo is good enough. I recommend Next.js framework to skip the boring stuff and start developing quickly. If you need the speed and/or low memory footprint try Rust and/or Go.

Links

https://drewdevault.com/2019/09/08/Enough-to-decide.html

JustDeleteMe in the real world

Mišo — Sat, 04 Jul 2020 16:21:09 +0000

Recently I was canceling some contracts with my internet provider, insurance, mobile operator, ... Unfortunately, it was not always easy and fast. I was wondering if there is some site for it. There are sites like JustDeleteMe with the guides on how to cancel virtual accounts, but I would like to see the same for the real contracts.

I know this is not so easy because the real world has borders and in every country, the rules are different even for the same company. But I hear from all sides how hard is to cancel some contract just because the company has some absurd rules, the intentionally complicated process, etc.

I heard that the EU plans to force organizations to not make canceling harder than it has to be. But it's not the thing today and the countries outside the EU will be probably out of luck.

Do you know such a site? Do you think it would be useful? I found https://www.cleevio.com/ but it's not the same thing I guess.