Setting Up a Project¶
For teaching purposes, I like to start people off using the TLA+ IDE. It abstracts out some parts of TLA+ that can be difficult for beginners. You can download the most recent version of the Toolbox here. You will need Java for it to work.
Once you have the toolbox, you’ll see an image like this:
Create a new specification under
File > Open Spec > Add New Spec.
Once you do that, you should see something like this:
---- MODULE wire ---- ====
For Historical Reasons, the
MODULE $name must be surrounded by at least four dashes, the module must end with at least four equal-signs, and the
$name of the module must match the filename (case sensitive). Everything above the module name and below the
==== is ignored, making those good places to store notes.
Let’s replace it with the contents of wire, so you get this:
(There should only be one
MODULE row at the top and one
==== row at the bottom. Make sure to change the name of the module in the first line if you named your spec
As mentioned in the core outline, we’re going to be teaching pluscal. Translate it under
File > Translate PlusCal Algorithm.
You can use the shortcut ctrl+T on Windows/Linux and cmd+T on Mac.
Once you do that, you should see this:
To actually check this with TLC, we have to create a new model to check. Do that under
TLC Model Checker > New Model.
Once you do that, you should see this page:
“What is the behavior spec” should be “Temporal Formula” and “Spec”. If it’s not, make sure you have only one set of
====in the spec, and the translated TLA+ is above it, then manually set the two fields.
Click the “Invariants” box to open it up.
Click “Add”, and then insert the text
Run the model, or press
When you run this, you will see an error pop up on the right side:
This is an error trace, showing the exact set of steps that lead to an invariant being violated. We’ll talk about it a bit more when we get into invariants in depth.
Making a Scratchfile¶
I often like to test the outputs of operators without having to run the entire spec. To do that, I have a separate spec that I call “scratch”:
---- MODULE scratch ---- EXTENDS Integers, TLC, Sequences Eval == 0 ====
This is different from a normal TLA+ file in two ways. First, instead of having “What is the behavior spec” set to “Temporal formula”, I have it set to “no behavior spec”. Second, on the “model checking results” page, I put
Eval in the “Evaluate Constant Expression” box.
Now whenever I run the model, the output of
Eval will be placed in the “Value” box below. In this case, it will be 0. But if I change the
Eval expression, I get something different.
- Eval == 0 + Eval == "hello world!"
Eval will put “hello world!”.
Having a scratchfile is very useful and I recommend setting one up. In the guide itself I will occasionally post “expression evaluations” like this
>>> 1+1 2
This just means that I set
Eval == 1+1 and got
2 as the output. You can use this to check that you got the same results as me.
And with that, we’re ready to start learning TLA+!