Re: [4diac-dev] InterfaceSpecBuilder

[Alois Zoitl <alois.zoitl@xxxxxx>]

hi,

I was never happy with the itnerfaceSpec parsing. It got a bit better with some helper methods we introduced but never really good. I find your approach very
nice and would be happy if you would like to contribute it to 4diac FORTE. 

Cheers,
Alois

On Mon, 2022-05-16 at 19:08 +0200, Davor Cihlar wrote:
> Hi!
> For iterating access to PostgreSQL result with multiple rows I went with a generic SIFBs (CGenFunctionBlock<CFunctionBlock>). I need some additional events
> and outputs and it doesn't make sense to do that through a comm layer.
> So I started writing my own GEN_PQ_ITER and everything was fine until I needed to fill out SFBInterfaceSpec. It's a nightmare to set everything up before
> filling it, and it's also not flexible at all. Further port changes are the same problems all over again + refactoring of the previous code.
> To simplify things I made SFBInterfaceSpec (source code attached). Does it make sense to incorporate it into Forte core? I also think that a parser for
> generic SIFBs should be standardized as currently every GEN_* implements its own parser.
> 
> An example would look like this:
> bool GEN_PQ_ITER::createInterfaceSpec(const char *pa_sConfig, SFBInterfaceSpec &pa_oInterfaceSpec) {
>     unsigned long numbers[2];
>     if (!parseSpec(pa_sConfig, "PQ_ITER", numbers, 2))
>         return false;
> 
>     static const std::array staticEINames = { g_nStringIdINIT, g_nStringIdREQ };
>     static const std::array staticEONames = { g_nStringIdINITO, g_nStringIdCNF };
> 
>     auto ifsb = CIfSpecBuilder();
>     bool stat =
>         ifsb.m_oEI.setStaticEvents(staticEINames) &&
>         ifsb.m_oEO.setStaticEvents(staticEONames) &&
>         ifsb.m_oDI.addData("QI", g_nStringIdBYTE) && /* g_nStringIdQI can be used instead of "QI" */
>         ifsb.m_oDI.addDataRange("SD_", (int)numbers[0]) &&
>         ifsb.m_oDO.addData("QO", g_nStringIdBYTE) &&
>         ifsb.m_oDO.addDataRange("RD_", (int)numbers[1]) &&
>         ifsb.bind(ifsb.m_oEI["INIT"], {ifsb.m_oDI["QI"], ifsb.m_oDI["ID"]}) &&
>         ifsb.bind(ifsb.m_oEI["REQ"], ifsb.m_oDI["QI"]) &&
>         ifsb.bind(ifsb.m_oEO["INITO"], ifsb.m_oDO["QO"]);
> 
>     stat = stat && ifsb.build(m_oMixedStorage, pa_oInterfaceSpec);
>     if (!stat) {
>         DEVLOG_ERROR("interface spec builder failed!\n");
>     }
> 
>     return stat;
> }
> As it can be seen, the builder supports both static and dynamic configuration (although not yet for "withs"). And in the end it uses only one member for data
> storage (m_oMixedStorage) which automatically cleans up at SIFB's EOL.
> I don't expect this method to use more memory than before. It even has potential to use less memory as there is only one memory block used (after setup). The
> only disadvantage would be slower initialization which shouldn't matter any way. By using this way of initialization in other GEN_* FBs the code size could
> also be reduced a bit.
> It would be nice to be able to use constexpr for std::array (so it can be stored in ROM), but something needs to be done first, I need to look into it.
> _______________________________________________
> 4diac-dev mailing list
> 4diac-dev@xxxxxxxxxxx
> To unsubscribe from this list, visit https://www.eclipse.org/mailman/listinfo/4diac-dev