This is the C++ core of the library. It can be used directly or wrapped with native-language glue. For example, for Android there is a Java (JNI) glue wrapper.

This library relies on the native environment to provide an HTTP request callback. Notes about its signature, inputs and outputs can be found in psicash.hpp.

There is an example implementation in the Android wrapper project. But note that it does not support proxied requests, which may be necessary depending on the environment. (E.g., it probably doesn't matter on iOS, since our app only supported full-device VPN. But on Windows the app mostly uses a local proxy, so the HTTP Requester must support proxying.)

Note that the requester must do HTTPS certificate validation.

All datastore reads and writes are mutexed. So, the only consistency guarantee is for individual data accesses. This means that multiple data accesses might get data from different states; for example, between getting the balance and getting the purchases list, there might have been a purchase, which would alter the balance. We will state this in positive terms as: "you will always get the very latest value". This is fine for our use cases at this time, but we might want to add "get a consistent set of data" in the future.

Platform-specific wrapper library implementations should not need additional synchronization. If more is needed, it should probably be added to the core library.

Mostly following Google's Style Guide. Major alteration: 4-space indent instead of 2. THe .clang-format file should be crafted as necessary to match what we want.

• I made no effort to be memory efficient (such as explicitly using move semantics, references, etc.). I tried to be clear and safe above efficiency. If there's a leak, that's a bug. If there's gross memory inefficiency, it can probably be improved. But we shouldn't bother with small efficiencies.

If Android Studio is saying that a new .cpp file is not part of the project, go to the "Build" menu and click "Refresh Linked C++ Projects". Then rebuild.

If you get a SIGABRT error in JNI code: You have probably triggered a JNI exception (that hasn't been cleared). It's possible that it's expected or acceptable and you just need to clear it, but it's more likely that it's a bug.

If you get a SIGSEGV error when hitting a breakpoint in JNI code: Yeah, beats me. I get it on MacOS but not Windows.

   +-----------+
   |           |
   | no tokens +<----------------------------+
   |           |                             |
   +-----+-----+                             |
         |                                   |
         +                                   |
 RefreshState+NewTracker                     |
         +                                   |
         |                                   |
         v                                   |
+--------+--------+                          +
|                 |                      ResetUser()
| Tracker tokens  |                          +
|                 |                          |
+--------+--------+                          |
         |                                   |
         +                                   |
       Login                                 |
         +                                   |
         v                             +-----+------------------------+
+--------+--------+    token expiry    |                              |
|                 +------------------->+  IsAccount() && !HasTokens() |
|  Account tokens |                    |                              |
|                 +------------------->+        "logged out"          |
+--------------+--+   AccountLogout()  |                              |
               ^                       +--+---------------------------+
               |                          |
               +--------------------------+
                     AccountLogin()