bash
cargo install -f fsm_gen
You can download from repository and build and install as usual
bash
cargo build --release
cargo install --path . -f
cpp generated code with option to move to error statelanguages, now we speak about templatesnamespace for functionsread template from file
Update idata in order to use ispush with btrees and more
Complete the cpp example and update on README.md
output with signals
Add templates
All computing processes consist of receiving an input and processing it by generating an output. From the simplest to the most complex.
Sometimes, the process depends on the context and it is necessary to manage a state.
Messages and state management, therefore, are two fundamental elements of any software process.
That's why I wrote two external DSLs long time ago for both elements (which I still use in production and are a great help).
In this repository I rewrite one of them, the code generator for a state machine.
At the moment it generates C++ code (my most immediate target in production).
You can have data (fields) on inputs structs, and also on each status
To explain the system, I will use the example in cpp_test/fsm
This example is about writing a system that will handle login requests.
First the server will be asked for a password.
This key will be used to encode the username and password (this one will be passed through a hash function) in the login request.
This encoding will be irreversible (hash function). The server will perform the same operation (starting from the hash of the password) to verify the validity.
If it is OK, it will send a login confirmation.
The diagram would look like this:
A list of transitions could be written as:
```peg // Example of fsm to manage login // on server side
[init] rqkey -> wlogin / send_key timer -> init
[wlogin] rqlogin -> login / sendlogin timer & timeout -> error -> wlogin [login] rqlogout -> logout / sendlogout heartbeat -> login / update_hb timer & timeout -> logout -> login
[logout] timer -> logout
[error] _ -> error ```
And this is the input for this tool to generate code
In fact, even the previous diagram has been generated from this DSL (it generated a graphviz dot file)
peg
[init]
...
init, w_login, login ... are the states
Depending on the input and the state (with its values as will be seen later), the system will change to a new state.
peg
rq_key -> w_login
If we receive an input (in this case rq_key) we go to next state (w_login)
The elements received by the status machine.
peg
INPUT
v
rq_key -> w_login
In the example they are rqkey, rqlogin, rq_logout, heartbeat and timer.
Functions that will be called depending on the status and input to decide the way forward.
peg
GUARD
v
rq_login & valid -> login / send_login
You can have more than one guard
peg
rq_login & valid & guard2 -> login / send_login
When you apply different options guards (or combination)...
In the example we have valid, timeout, ontime.
peg
rq_login
& valid -> login / send_login
-> logout / log_err
It could be written with negative guard
peg
rq_login
& !valid -> logout / log_err
-> login / send_login
Behind the -> arrow is the state we will change to.
peg
FINAL_STATUS
v
rq_key -> w_login
We can define an action to be performed when executing a transaction.
This will be after the final state and '/'.
peg
ACTION
v
rq_login -> logout / log_err
You can have more than one action
peg
rq_login -> logout / log_err action2
In this example we have sendkey, sendlogin...
In all states it is necessary to consider all inputs.
But it is very common that many transitions are the same (generally error cases).
This is marked with the input _
Consider the this example status:
peg
[init]
rq_key -> w_login / send_key
timer -> init
_ -> logout / log_err
_ means... any other input...
Therefore considering all possible inputs in this state
error status and implicit transitionserror is a special status
You can move to error status explicitly.
Any transition no defined, will finis on error status.
You can also put some verifications on transiction funcion, and incase
of fail, you can move to error (even when is not explicitly writted on fsm)
This is so, because checking the params, is so commont that adding guards for it, would generate a lot of sound
In our example...
peg
[init]
rq_key -> w_login / send_key
timer -> init
There are no transations for rq_login and rq_logout. Both are implicit and is
equivalent to...
peg
[init]
rq_key -> w_login / send_key
timer -> init
_ -> error
This is the transition function control (obiosly you can specialize as you wish)
cpp
// status change functions
template <typename FROM, typename IN, typename TO>
std::variant<TO, st_error_t> fromin2(const FROM &, const IN &) {
// here you could check the params and decide to go to error
// instead to the programmed trasnsition
...
}
And this is when you explicitly write a transition who finished on error status
cpp
template <typename FROM, typename IN>
st_error_t fromin2error(const FROM &, const IN &) {
return st_error_t{...};
}
Another way to end on error transition. If an exception is thrown while the input is being procesed...
```cpp [...] } catch (...) {}
auto nwstinfo = fromin2error
Comments starts at // and continues to the end of line
To get help...
bash
fsm_gen --help
```txt
fsm_gen -h
fsm_gen 0.6.1 jleahred
Generate code from a simple fsm file
To check the supported templates --show_templs
USAGE: fsmgen [FLAGS] [OPTIONS] [fsmfiles]...
FLAGS: -d, --dot-graphviz Generate graphviz dot file -h, --help Prints help information --help-cpp Give me some information about generating cpp files -s, --show-templs Show supported template generators -V, --version Prints version information
OPTIONS:
-T, --threads
ARGS:
The default template is c++ (and at the moment the only one)
You can run:
bash
fsm_gen login.fsm
And it will generate the c++
You can pass a list of fsm files
bash
fsm_gen login.fsm test/seller.fsm test/test2/lift.fsm
The code will be generated on same directory of original .fsm file
If your shell supports it, you could run...
bash
fsm_gen **/*.fsm
Starting from the example login.fsm the system will create...
txt
fsm_login_gen.h
fsm_login_gen.cpp
You don't have to modify these files.
You have to write your code on next files...
txt
fsm_login_types.h
fsm_login_private.hpp
These two files will be created if don't exist as a reference
A empty reference code for these two files will be added on _gen files as comments.
Files dependency:
Full code on cpptest/fsm/fsmlogin_gen.h
We are informed on when it was created.
This file is the starting point to use or extend the generated status machine in your program.
cpp
// generated automatically 2019-03-22 11:24:40
// do not modify it manually
Headers and namespaces based on filename
```cpp
namespace login { ```
Internal class forward declaration
cpp
class BaseState;
typedef std::shared_ptr<BaseState> SState;
The fsm class you have to instantiate or extend
```cpp // ------------------- // F S M class Fsm { public: Fsm(); ~Fsm();
void process(const heartbeatt& in); void process(const rqkeyt& in); void process(const rqlogint& in); void process(const rqlogoutt& in); void process(const timert& in);
... } ```
Full code on cpptest/fsm/fsmlogin_gen.cpp
You must not modify this file, and it's not necessary to know much about it
Full code on cpptest/fsm/fsmlogin_types.h
In this file you have to declare the status info types, and input types
If the file doesn't exist, it will be created with empty data types
cpp
// Code generated automatically to be filled manually
// This file will not be updated by generator
// It's created just the first time as a reference
// Generator will allways create a .reference file to help with
// new methods and so
On namespace based on fsm file, we have the two types to declare
You can, of course, complete these types on cpp file if necessary
```cpp namespace login {
// status info types struct stinitt{}; struct stwlogint{}; struct stlogint{}; struct stlogoutt{}; struct sterror_t{};
// input types struct inheartbeatt {}; struct inrqkeyt {}; struct inrqlogint {}; struct inrqlogoutt {}; struct intimer_t {};
} // namespace login
```
Full code on cpptest/fsm/fsmlogin_private.hpp
This is the other file you have to maintain by hand.
```cpp
// Code generated automatically to be filled manually // This file will not be updated by generator // It's created just the first time as a reference // Generator will allways create a .reference file to help with // new methods and so
// This file will be included in _gen.cpp // (anywhere else)
// to make happy some IDEs
namespace { using namespace login; ```
As you can see, it is on anonymous namespace.
This file is private, it will be included by _gen.cpp. Defining an anonymous
namespace, we keep this implementation as private. Even more important, the compiler
will alert us if we forget an implementation, also if one is not necessary.
Log and status change transitions are templates.
This is a log example:
cfg
[init] rq_key -> w_login
[w_login] rq_login(valid) -> login
[login] rq_logout -> logout
In this way, you can specialize or generalize as much as you want.
```cpp
// log
template
// status change functions
template <typename FROM, typename IN, typename TO>
std::variant<TO, st_error_t> fromin2(const FROM &, const IN &) {
// you can specialize this generic function
return TO{};
}
template <typename FROM, typename IN>
st_error_t fromin2error(const FROM &, const IN &) {
return st_error_t{};
}
```
First parameter in log, is an string with transition change information (initial transition, input, guard if so, final transition)
Next, we have the guards and actions functions.
```cpp // guards bool valid(const inrqlogint& /*in*/, const stwlogint& /st_info/) { return true; } bool timeout(const intimert& /in/, const stwlogint& /*stinfo/) { return true; } bool timeout(const in_timer_t& /in/, const st_login_t& /st_info*/) { return true; }
// actions
void act_send_key(const st_init_t& /*st_orig*/, const in_rq_key_t& /*in*/, const st_w_login_t& /*st_dest*/) {}
void act_send_login(const st_w_login_t& /*st_orig*/, const in_rq_login_t& /*in*/, const st_login_t& /*st_dest*/) {}
void act_send_logout(const st_login_t& /*st_orig*/, const in_rq_logout_t& /*in*/, const st_logout_t& /*st_dest*/) {}
void act_update_hb(const st_login_t& /*st_orig*/, const in_heartbeat_t& /*in*/, const st_login_t& /*st_dest*/) {}
} // namespace anonymous ```
```dot digraph G { rankdir = BT; node [shape=plaintext fontname="Arial"];
main -> "fsmlogingen.h"
"fsmlogingen.cpp" -> "fsmlogingen.h"
"fsmlogingen.h" -> "fsmlogintypes.h"
} ```