Michael Teter

LiveBook Can't Connect After Upgrade

Wed, 07 Aug 2024 22:18:00 +0000

After LiveBook minor version upgrade, can’t connect to node

There was a “small” change in the recent LiveBook project which may prevent LiveBook from connecting to other nodes.

LiveBook Changelog

To make your node visible to LiveBook now, you must use long names.

Now when starting iex, use ✅ --name instead of ❌ --sname.

Then in the LiveBook config, specify the full node name such as myproject@mymac.local (for local Mac users).

Macos + Wezterm + VSCode light/dark switcher

Thu, 25 Jul 2024 19:36:00 +0000

My zsh command-line approach to toggling my light/dark modes for the dev tools I use

In my ~/.zshrc I have the following:

# Function to change WezTerm, VSCode theme, and macOS appearance
change_themes() {
  local wezterm_theme=$1
  local vscode_theme=$2
  local macos_mode=$3
  local settings_file="$HOME/Library/Application Support/Code/User/settings.json"

  # Update WezTerm theme
  sed -i '' "s/^config\.color_scheme.*$/config.color_scheme = \"$wezterm_theme\"/" ~/.config/wezterm/wezterm.lua

  # Update VSCode theme
  sed -i '' "s/\"workbench\.colorTheme\": \".*\"/\"workbench.colorTheme\": \"$vscode_theme\"/" "$settings_file"

  # Update macOS appearance
  if [ "$macos_mode" = "dark" ]; then
    osascript -e 'tell app "System Events" to tell appearance preferences to set dark mode to true'
  else
    osascript -e 'tell app "System Events" to tell appearance preferences to set dark mode to false'
  fi
}

alias light='change_themes "Alabaster" "Default Light+" "light"'
alias dark='change_themes "3024 (dark) (terminal.sexy)" "Panda Syntax" "dark"'

Then from any terminal app that’s using zshell and my normal zshell dotfile,

light to change my themes to Light mode.

dark to change the themes to Dark mode.

Obviously you can choose your own theme names in the aliases, and you can change the alias names as well. (Be sure to check which dark or whatever names you choose so you don’t shadow some other program with that name.)

Remember to source ~/.zshrc after you change your zshell config file so your changes are recognized in your current shell while setting this up.

When to use halt() in Phoenix Framework

Sat, 19 Aug 2023 17:06:00 +0000

TLDR;

In a pipeline function, call halt() after a redirect().

In a controller action, it is not necessary to halt after a redirect.

Halt - The Confusion

If you’re a Phoenix beginner like me, you may have seen some code where a redirect() is followed by a halt().

The code generated by mix phx.gen.auth includes a module called UserAuth, which has the following function defined:

def redirect_if_user_is_authenticated(conn, _opts) do
  if conn.assigns[:current_user] do
    conn
    |> redirect(to: signed_in_path(conn))
    |> halt()
  else
    conn
  end
end

But you can also find examples where redirect is not followed by halt. UserSessionController has the following function:

defp create(conn, %{"user" => user_params}, info) do
  %{"email" => email, "password" => password} = user_params

  if user = Accounts.get_user_by_email_and_password(email, password) do
    conn
    |> put_flash(:info, info)
    |> UserAuth.log_in_user(user, user_params)
  else
    conn
    |> put_flash(:error, "Invalid email or password")
    |> put_flash(:email, String.slice(email, 0, 160))
    |> redirect(to: ~p"/users/log_in")
  end
end

The Phoenix Controller docs docs illustrate a redirect followed by a halt, and the text below their example states that Plug.Conn.halt/1 is “by default imported into controllers”.

These different examples can make one wonder when and why to use halt().

What does halt do?

The docs for halt(conn) are brief, saying it “Halts the Plug pipeline by preventing further plugs downstream from being invoked.”

Phoenix is built around the concept of transforming some input data through a pipeline (a series of pipe stages), resulting in output data which is then returned to the client. But when circumstances dictate a change in paths (such as with a redirect), then some pipelines need to be stopped.

Looking in the Router module, we can see something like this:

pipeline :browser do
  plug :accepts, ["html"]
  plug :fetch_session
  plug :fetch_live_flash
  plug :put_root_layout, html: {ExampleWeb.Layouts, :root}
  plug :protect_from_forgery
  plug :put_secure_browser_headers
  plug :fetch_current_user
end

Each of these plugs gets called in sequence, with the output of the previous stage being the input of the next stage. Any one of these stages could decide that the process needed to deviate or stop based on any number of reasons.

As we see earlier, redirect_if_user_is_authenticated() branches based on a condition in the current state (which is usually called conn). In one path, it redirects and halts. In the other path, it simply returns appropriately so the next stage of the pipeline can be executed.

In this case, the halt is important because it tells Phoenix to stop further processing of this pipeline. Meanwhile, the redirect will have resulted in another pipeline running, and that pipeline should be the one to determine and return the final data.

When to use halt

USE

Call halt() when you need to stop the pipeline from performing any following stages. This would be in a plug function after a redirect.

DON’T USE

Since a controller action is the last stage of a pipeline, there is no need to call it within the action (even if you have a redirect). It is harmless if used at the end of the pipeline, however…

Why was that confusing?

Perhaps it’s just me, but I was confused about this probably mostly because of the previously mentioned line in the controller docs which stated that halt “is by default imported into controllers”. Why would it be automatically available to controllers if controller actions are the last step of a pipeline?

The reason is that you may wish to write some function plugs in a controller (similar to Rails’ before filters) which would be run before your actions in that controller. And your custom plug might wish to stop the current process and redirect to a new one. Therefore, your plug function would need to use halt.

I still find the choice to make halt automatically available in controllers curious, because it implies that there would be so many times you would need it that having it automatically included was a significant convenience. The implication is that a controller will have more plug functions (with redirects) than actions (which don’t need to halt even if they redirect). I kind of doubt that’s the reality.

I would expect controllers to have more actions than plug functions which redirect. And in that case, I think it would be clearer to make the developer explicitly import halt/1 when needed.

Wordpress (on dynamic IP virtual machine) redirects to old IP

Sat, 29 Sep 2018 13:42:00 +0000

Problem

Wordpress on a Virtual Machine keeps redirecting to old IP address

Using a virtual machine with a dynamic IP bridged to the host machine, and the host machine connects to various different WIFI networks… I found that my development Wordpress site on my (virtualbox) VM kept redirecting my web browser to the old/original IP address the VM had when Wordpress was first setup.

I do not yet know the correct solution; surely there is one, but the crude manual fix is to change the (unfortunately?) hard-coded IP address in the Wordpress site database.

Solution

mysql> select * from wp_options where option_name in ('siteurl', 'home');       
+-----------+-------------+---------------------+----------+
| option_id | option_name | option_value        | autoload |
+-----------+-------------+---------------------+----------+
|         2 | home        | http://192.168.1.17 | yes      |
|         1 | siteurl     | http://192.168.1.17 | yes      |
+-----------+-------------+---------------------+----------+

You’ll probably see two rows returned that have the VM’s old IP address. (What were they thinking with this?)

Surely this is a dirty fix, and there’s a more correct solution. But make the following change:

mysql> update wp_options set option_value='http://10.0.1.8' where option_name in ('siteurl','home');
Query OK, 2 rows affected (0.00 sec)
Rows matched: 2  Changed: 2  Warnings: 0

With this change, the incorrect redirect should be resolved.

Naturally you will use whatever IP address in the sql update that your current VM is using; and pay attention to your original option_value. If it was https, then use https in your SQL update.

Compiling Elixir Modules In Windows

Fri, 06 Jul 2018 08:55:00 +0000

Problem

I wanted to do some Elixir/Phoenix development in Windows 10 (don’t ask why…), but I couldn’t build my Phoenix app due to a failure building the comeonin module.

could not compile dependency :comeonin, "mix compile" failed. You can 
recompile this dependency with "mix deps.compile comeonin", update it with "mix
deps.update comeonin" or clean it with "mix deps.clean comeonin"
** (Mix.Error) Could not find the program `nmake`.

The issue is that even though I had installed the Visual Studio Community development tools (which includes nmake), the build tools are not visible in the path on Powershell.

Visual Studio does provide a batch file to make those tools visible, but running that script only makes the tools visible to the (temporary) shell that the script runs in.

Solution

Pre-reqs

Erlang and Elixir (Windows packages) are installed.
MS Visual Studio, or the Community version, or the MS build tools, are installed.

Build Tools Paths

From pre-req 2, determine the path to your vcvars64.bat script. For me this was

C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Auxiliary\Build\vcvars64.bat

Powershell Profile

Edit the Powershell profile file. If the file doesn’t exist, create it. The path is

C:\Users\your_username_here\Documents\WindowsPowerShell\Microsoft.PowerShell_profile.ps1

Add the following code to the profile

function Invoke-CmdScript {
  param(
    [String] $scriptName 
  )
  $cmdLine = """$scriptName"" $args & set"
  & $Env:SystemRoot\system32\cmd.exe /c $cmdLine |
  select-string '^([^=]*)=(.*)$' | foreach-object {
    $varName = $_.Matches[0].Groups[1].Value
    $varValue = $_.Matches[0].Groups[2].Value
    set-item Env:$varName $varValue
  }
}

Invoke-CmdScript "C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Auxiliary\Build\vcvars64.bat"

The function allows you to run a script, the one that makes the build tools visible, but instead of losing the context after that script ends, the settings are propogated back to the calling shell (Powershell).

The profile then uses that Invoke-CmdScript function to call the vcvars64.bat script that we got from step 2 of the prereqs.

With this done, you should be able to successfully build/compile Elixir modules within Windows 10.

Solving the Elixir/Ecto error 'undefined function from/2'

Tue, 06 Mar 2018 03:17:00 +0000

Problem

In Elixir, when trying to write an Ecto query such as this:

q = from(v in Vol,
    where: v.user_id == ^user.id,
    order_by: [desc: v.inserted_at])

you may encounter the error:

undefined function from/2

This is a rather difficult error to search for on a search engine, and more than once over time I have encountered it and had to dig around to find the solution.

Solution

import Ecto.Query

Do this before your use of from/2 to include the functions provided by Ecto.Query.

Add Logo to All Images In Folder

Mon, 06 Mar 2017 01:07:00 +0000

Goal

We want to superimpose (or overlay) our logo image on top of all images in the current directory.

These instructions are for macOS, and they assume a basic level of familiarity with entering commands via the Terminal.

This example shows how to overlay an image (such as a logo) on every image in the current directory. It works on macOS 10.12, and it probably works in Linux.

The command will create new images having the original image name with the word logo_ prepended. So an_image.png is used to create a new file, logo_an_image.png.

Pre-requisites

You’re on a modern Mac
You have HomeBrew or MacPorts installed
You have installed ImageMagick

Assumptions

You aren’t afraid to use the Terminal
You want to process all images in the current directory
Your logo image is in a subfolder of this directory called “logo”, and the logo filename is “logo.png”

You can change all of the settings and names once you understand the command.

Risks and Caveats

If something goes wrong, you could lose data. It is wise to make regular backups, and it is also wise play with this command first in a test folder with some sample images and logo image.

This may not be the most efficient or terse solution to the problem, so feel free to share better solutions.

cd my_test_folder_with_copies_of_images

find . -type f -maxdepth 1 \( -iname "*.jpg" -o -iname "*.png" \) \
    | tr -d './' \
    | tr '\n' '\0' \
    | xargs -0 -n1 -I {} \
    composite -gravity SouthWest -geometry +10+10 \
        logo/logo.png {} logo_{}

What is Happening?

In short, we get a list of the image files in the current directory, then use the composite command from ImageMagick to add our logo image down near the lower left (SouthWest) corner of each image from our found list.

Let’s explore this line by line. Note that the trailing \ (backslash) tells the shell to continue the command on the next line. it’s a good way to break up the long compound command into more readable chunks.

find . -type f -maxdepth 1 \( -iname "*.jpg" -o -iname "*.png" \) \

First, we need to get a list of all image files in the current directory. The command for that is find.

. (dot) means apply the command to the current directory
-type f means to only list files (not directories)
-maxdepth 1 means to only look at the current level; don’t go into any sub-directories
\( -iname "*.jpg" -o -iname "*.png" \) from the inside out, this says to use the -iname flag to search, case-insensitive, for files ending in .jpg, -o (or) (do another case-insensitive search for files ending in .png). Then we need to wrap that expression with ( and ) so they get processed together. If we don’t do that, the arguments will get parsed from left to right in a way that will not give us the correct result. Lastly, we have to escape the ( and ) with leading \ backslashes so the shell doesn’t try to interpret them.

Next, we pipe | the output of find into sed, the stream editor. find is listing the files it finds with a path showing the current directory (.) - it looks like this: ./an_image.png. However, since we will need to prepend the word logo_ to the beginning of our output files, we need to get rid of the ./ from each source file name. sed does this for us.

provide a substitution rule string, denoted by the leading and closing ' (single quote)
s|a|b| says “search” for a, replace with b
our a in this case is \./ , which is the ./ that we want to remove, but we have to escape that . so the shell won’t try to interpret it
the b (replace) field is empty because we don’t want to replace the search text with anything - we just want to eliminate the search term

Then we pipe the output of the previous step into tr, translate.

Case-insensitive Sorting in Swift 3

Sun, 11 Sep 2016 22:34:50 +0000

Sorting with CoreData is pretty simple, but I ran into difficulty while trying to get string sorting to be case-insensitive. Examples I found worked, but they gave deprecation warnings regarding using selectors. This is the solution that worked for me:

sortDescriptors.append(
  NSSortDescriptor(key: "title",
                   ascending: true,
                   selector: #selector(NSString.localizedCaseInsensitiveCompare)))

Of course this leaves out a lot of detail, but the importance of the example is in the proper (Swift 3) selector use.