[Bug] import tgis takes too long

[petrasovaa]

Describe the bug

Importing tgis takes a long time, almost 3 seconds. We need to first investigate why and if there is anything to do about it.

[echoix]

I already imagined creating a "lite" c-library that only exposes common, useful (for Python) api surface, so the generated bindings are smaller. Advanced features available on demand in another file.

But that's only part of a solution; another part is the whole work done on initialisation (importing the Python code, that ends up importing almost all of temporal and more). (Like DB connections and all, iterating through lists or data structures probably too much).

I'm pretty sure something is slowing us down or almost deadlocking us in the rpc mechanism, but I never managed to profile it successfully and conclusively, most of my attempts had results that I didn't trust.

If I would have had a real observation and result, I would have done it long time ago, in my dream of not having temporal test take the same time as the rest of all grass tests.

[ninsbl]

Yes, .
For modules that create STDS (where C-funtions are used), 3 seconds import is probably neglectible overhead.

For tools that just read data from the TGIS DB that is probably different (and here no C-functionality is needed).

I think tests are yet another issue. One (and probably the main) reason for slow tests is that test-data is almost always created first and more data (n>1 maps) is needed...

[petrasovaa]

This is Linux specific, mac doesn't have this problem (no idea about Windows). Potential fix: #7605

[echoix]

Windows feels slower with everything.