Detailed SDL tutorial
- 1 OpenGEODE - SDL tutorial
- 2 Introduction
- 3 SDL concepts overview
- 3.1 Start symbol
- 3.2 State / Nextstate
- 3.3 Input
- 3.4 Continuous signals
- 3.5 Connection
- 3.6 Output
- 3.7 TASK
- 3.8 Procedure calls
- 3.9 Decisions
- 3.10 Labels and branches
- 3.11 Procedures
- 3.12 Composite states
- 3.13 SDL Textual grammar
- 4 SDL in the context of TASTE
OpenGEODE - SDL tutorial
Author: Maxime Perrotin
contact: maxime (dot) perrotin (at) esa.int
SDL is a rich language and the complete specifications are available on the ITU-T website : https://www.itu.int/rec/T-REC-Z.100
There are several major revisions of the language:
- SDL88 - the first public version
- SDL92 - major update adding object orientation
- SDL96 - minor update fixing issues of SDL92
- SDL2000 - major update introducing new concepts (agents, exceptions, parallel and nested states)
- SDL2010 - the baseline of the current version (latest version is from 2019)
Check https://www.sdl-forum.org for more information.
OpenGEODE supports a subset of the last version of the standard. In this page we list the features of this subset and explain how to use them.
SDL concepts overview
These are the main symbols used in a SDL state machine:
State machines in SDL look like flow diagrams and are read vertically. This is a typical SDL transition diagram:
We will now go in the details of each symbol.
There are three places in a SDL model where a START symbol can be used:
1. At the state machine root diagram: It is mandatory here. A state machine has exactly one start transition.
The start transition is executed at process creation
The start transition
- Sets the initial state - May execute initial actions (initialization of variables)
2. Inside procedures. The start symbol has a slightly different shape, and is triggered when the procedure is called:
3. Inside nested states
There can be more than one Start transition in a nested state. In that case they must be given a name.
The selection of the nested state start transition is done one level above, when entering the nested state, using the via syntax:
State / Nextstate
- Each state has a name
- In a given state, the process is expecting to receive messages
- A state can be composite
Note the following shortcut that allow to save diagram space:
A new state is usually reached at the end of a transition:
A shortcut is the history state, that returns to the most recent state. This is particularly useful when combined with the state shortcuts:
If the most recent state was a nested state, re-entering the nested state means going through its startup transition and possibly executing its entry procedure.
It is also possible to enter a nested state via one of its internal startup transition, as seen above using the via syntax.
When a state is an aggregation (i.e. containing parallel states) it is possible to consume a transition at the parent level and return to the parallel states in the same state they were before the transition. This is called "history dash", and is achieved using the identifier -* in the next state symbol:
- Fires a transition : the transition is executed when the process consumes the signal
- In a given state, the process can expect several signals
- May have parameters (use variables to store their values)
The following shortcuts are available:
- Inputs at level N have priority over inputs at level N-1 (composite states)
- As a consequence, be careful with « asterisk » inputs : if the state is composite, all inner inputs are ignored.
Continuous signals are spontaneous transitions that trigger when no input signal is present (standard input always have priority). They are boolean expressions:
These can be used to perform background tasks, that will be interrupted as soon as a message is received.
Continuous signal should be mutually exclusive otherwise there is a case of non-determinism and the order of evaluation is not guaranteed (do not assume they are evaluated from left to right).
It is possible to use a priority to express the order of evaluation:
This will be properly reflected if you generate the code:
A connection is a transition that is executed when leaving a nested state. It is named in the nested state itself using the exit symbol, like this:
The connection symbol, one level above, allows to trigger a corresponding transition when leaving the state:
Nested states also contain optional entry and exit procedures, that are executed in addition to the start transition and to the connection transitions. See below for details.
Transmission (sending) of a message, with or without a parameter.
NOTE: only one parameter is supported by TASTE.
Tasks are elementary actions of a process transition. They can be either informal:
(use single quotes and free text)
Formal tasks allow to assign values to variables and to execute for loops to iterate on arrays. It is also possible the manipulate data with built-in operators (ternary, math, array concatenation, etc.). The details are given on this page:
- have in and in/out parameters
- return a variable
- be defined locally or externally
The procedure call symbol is for calling procedures that do not return variables. Parameters, if any, are separated with commas.
There are two special built-in procedures that can be called to write strings on screen: write and writeln. They can take multiple parameters separated with commas, for example to display variables of basic types. For exmaple:
Procedures that return a value must be called within a TASK symbol using this syntax:
There are also built-in procedures for setting and resetting timers: set_timer(time_in_ms, timer_name) and reset_timer(timer_name).
Like tasks, decisions can be formal or informal. Informal decisions are useful when building the model, and can be made formal when ready.
Informal decision and their answers are expressed with single quotes:
While formal decisions contain expressions:
A decision can have more than two answers, and in that case the answers must be mutually exclusive. The last answer can be the special kewword else. Decisions can be used to build graphical loops, when used together with join/labels:
If the expression in the decision is an ENUMERATED variable, the decision answers can be the enumerants:
If the expression is the built-in present operator for a CHOICE type variable, the answers are the possible elements of the union:
There is a special decision in SDL called decision any that allows to trigger a branch randomly:
It is also possible to group answers in convenient ways to limit the number of branches:
Labels and branches
Labels and branches allow re-routing and basic loops. They can be used to go to a next state from multiple points without repeating common pre-entry actions (as an alternative to a state entry procedure).
Procedure are sequential sub-functions. They can have in and in/out parameters and optionally return a value. They have visibility on the parent variables. They can contain local variables, but no internal state (yet standard SDL allows it).
The symbol for adding a procedure to the SDL model is the following:
If the procedure has parameters, they must be declared inside the procedure in a text box with the following syntax:
Procedures can have special attributes, in which case the declaration of the procedures have to be done in a text box at a higher level (not in the procedure itself).
A procedure can be declared external. This means that the definition of the procedure is not part of the model. When generating code that implies that the user (or TASTE) provides the code of the procedure and a symbol for the linker.
A procedure can be declared at one place of the model, and implemented (with the graphical symbol) at another place. In that case the declaration must contain the keyword referenced.
Exported procedures allow to implement the concept of remote procedure calls. When a procedure is declared as exported, it means that it can be called synchronously from the environment of the process without an explicit signal. When targeting the generation of code, this is very convenient for example to implement getters or setters that can run in the context of the calling thread and that do not require a pair of signals for sending back a value.
Exported procedures in SDL are declared like this:
They must then be implemented graphically and the procedure content must contain a text box with the same interface declaration:
As you can see in this example the procedure has visibility on the parameters, but also on the containing process context (dataHasChanged is a variable declared at process level).
After the execution of the exported procedure, the containing process will execute an optional transition with the same name as the procedure (but with no parameters). This allows to change the state of the state machine if needed. For example:
This does not require a signal declaration, as it is implicit from the procedure declaration.
If there is no transition defined, it is also possible to trigger a transition using continuous signals. For example:
OpenGEODE supports both nested and parallel states. Double click on a state to create a nesting structure.
If the neseting structure only contain states but no start transitions, they are parallel states. Each of them must then be refined and have their start transitions.
Parallel states can't consume the same signals (signal lists are disjoint).
A nested state can have multiple entry and exit points, as well as an entry and exit procedures which are called automatically upon entering/leaving the nested state.
Nested state entry points
Inside a nested state you can define multiple startup transitions. In that can they must be named:
From the state above, you can choose which startup transition you want to use, using the via clause:
You can add an unnamed startup symbol for the default entry without a via clause.
Nested state exit transitions
Similar to the the startup transition you can define named exit:
In the level above, use the CONNECT symbol :
and specify the exit labels you defined inside the state:
entry and exit procedures
Inside a nested state you can define a procedure with the special names entry and exit
These procedures will be called automatically when the state is entered or left, from all defined paths.
SDL Textual grammar
SDL is both a graphical and a textual language. OpenGEODE's SDL grammar is defined here:
SDL in the context of TASTE
One important features of SDL is the possibility to describe a system made of components that communicate through messages. This description can be nested: a block can contain other blocks that eventually contain actual state machines.
This is not directly supported by OpenGEODE because it is done in TASTE (graphically and textually) using the AADL language.
The semantics are nearly similar, with the following differences:
1. SDL does not allow to specify a cyclic message in this view (periodic activation has to be done using timers inside state machines)
However TASTE allows it in the Interface View:
In this example, the interface named "monitor" is cyclic. A period has to be specified for it.
2. In regular SDL all messages are asynchronous. Direct function calls are possible between two state machines (remote procedure calls) but this communication is hidden from the diagram.
In TASTE, synchronous calls are expressed in the Interface View to expose the remote procedure calls from SDL to an external function.
Synchronous calls are immediately executed (blocking calls) and can be either protected (mutual exclusion between messages) or unprotected (executed immediately no matter what).
TASTE supports synchronous interfaces:
- Synchronous provided interfaces are implemented using the declaration of exported procedures.
- Synchronous required interfaces are implemented using the declaration of external procedures.
Both constructs are part of the SDL standard.
3. In SDL all active functions are state machines
In TASTE it is possible to implement them in different languages: SDL, but also Simulink, C, C++, Ada, and even VHDL. TASTE generates the glue code between the functions.
4. In SDL, communication is done via messages (called signals) that are defined at system level. This means that a signal name is unique across the system.
When sending a message, it is possible in SDL to specify the recipient in case several can receive the same signal name. Broadcast and multicast are also supported.
In TASTE, the interfaces are defined at function level. Two functions can therefore have the same interface name but with different semantics (different parameters). It is possible to rename the interface at the sender side to avoid ambiguities.
While SDL offers the possibility to specify the recipient when sending a message (OUTPUT message TO sender) there are two things to consider:
a. the state machine's code has to be aware of the system it is connected to, making reuse more difficult in some cases
b. other languages such as C or Ada do not offer such construct
This is how the problem is addressed in TASTE:
5. SDL allows to specify multiple parameters associated to asynchronous signals.
In TASTE however, it is possible to have only one parameter in asynchronous interfaces (one signal = one message). Synchronous interfaces support multiple in or in/out signals and are supported by TASTE and OpenGEODE.
6. The SDL standard comes with two ways to describe data types: a legacy (yet very powerful) type system, and ASN.1 TASTE only supports a very small subset of the legacy SDL type system, and relies on ASN.1 instead. ASN.1 is an international standard (ISO and ITU-T), supported by multiple tools and used in lots of applications. It is therefore recommended to use it instead of the built-in SDL syntax for data types.