Hi Adam, OK, but would I need to know the fields in the message ahead of time? That is what I am trying to avoid. The paths of the nested messages are in a table, such as
"healthreport/os/version", 3 So the healthreport variable is allocated, and it contains a component called os, which itself contains a field called version. So I am trying to write a generic enum function, that can descend the parent, and guided by the path info, can resolve the value that needs to be stored there, but if I am to traverse the entire tree manually, how does the reflection API help? On Tuesday, 22 June 2021 at 14:21:59 UTC-4 [email protected] wrote: > I think the reason this is getting tricky is because you're trying to > traverse the descriptors first and then look at the message tree afterward. > I would expect it to be much easier if you traverse the message and look at > the descriptors at the same time. > > On Tue, Jun 22, 2021 at 10:58 AM J G <[email protected]> wrote: > >> Hi Adam, >> >> That works for the first iteration, but I descend the tree like so: >> >> bool enumpb( const char * pszpath, ENUMPROTOPROC f, const >> google::protobuf::Descriptor * d, uintptr_t param ) { >> >> std::string path = pszpath; >> >> for ( int i = 0; i < d->field_count(); i++ ) { >> >> auto field = d->field( i ); >> >> std::string localpath = pszpath; >> >> if ( 0 != strcmp( "component", field->name().c_str() ) ) { >> >> localpath.append( field->name() ); >> >> } >> >> if ( ( ! localpath.empty() ) && ( '/' != localpath.back() ) ) { >> >> if ( f && ! f( field, localpath.c_str(), param ) ) { >> >> return false; >> >> } >> >> } >> >> auto mt = field->message_type(); >> >> if ( ! mt ) { >> >> continue; >> >> } else if ( 0 != strcmp( d->full_name().c_str(), >> mt->full_name().c_str() ) ) { >> >> std::string localpath2 = localpath; >> >> if ( ( ! localpath.empty() ) && ( '/' != localpath.back() ) ) >> { >> >> localpath2.append( "/" ); >> >> } >> >> if ( ! enumpb( hp, report, localpath2.c_str(), f, mt, param ) >> ) { >> >> return false; >> >> } >> >> } else { >> >> // printf( "Skipping circular %s" EOL, d->full_name().c_str() >> ); >> >> } >> >> } >> >> return true; >> >> } >> >> So I start the traversal like this: >> >> auto d = report->GetDescriptor(); >> >> enumpb( "", f, d, param ); >> >> And it goes down the variable, visiting each leaf and nested child >> variable, but I can't address each nested child directly that way, can I? >> The mt variable does hold the descriptor for each nested variable at some >> point, but I don't know how I'd derive the variable's instance from it. >> >> On Tuesday, 22 June 2021 at 13:47:15 UTC-4 [email protected] wrote: >> >>> I think the easiest thing would be that wherever you're now storing a >>> google::protobuf::FieldDescriptor*, you can also store a >>> google::protobuf::Message* pointing to the parent message. >>> >>> On Tue, Jun 22, 2021 at 10:41 AM J G <[email protected]> wrote: >>> >>>> Hi Adam, >>>> >>>> Yes, the HealthReport variable is the parent, and it contains a >>>> HardwareComponent variable, but I am enumerating the from the parent, >>>> meaning I am trying to not hard-code the structure of the contained items. >>>> >>>> So how would I obtain a pointer to the message for each leaf without >>>> hard-coding the member names in there? >>>> >>>> I am able to figure out what value I want to set in each leaf by a map >>>> I have that uses the field's path to match it to the value I want to store. >>>> >>>> On Tuesday, 22 June 2021 at 13:33:11 UTC-4 [email protected] wrote: >>>> >>>>> So is it correct that HealthReport is the top-level message type and >>>>> HardwareComponent is nested somewhere within that? I think what you're >>>>> trying to do is doable, but when you call reflection->SetString(), you >>>>> have >>>>> to pass the immediate parent message containing the field, not the >>>>> top-level message. You don't need to save the descriptor for each leaf, >>>>> but >>>>> you do need to save a pointer to the message containing each leaf. >>>>> >>>>> On Tue, Jun 22, 2021 at 10:17 AM J G <[email protected]> wrote: >>>>> >>>>>> Hi again Adam, and thank you for taking the time to help me. >>>>>> >>>>>> Maybe I haven't explained what I am trying to do properly. >>>>>> >>>>>> I have a protobuf variable, which itself is composed of more nested >>>>>> variables. >>>>>> >>>>>> I am enumerating the fields of the variable. >>>>>> >>>>>> Where the item is a leaf, the field is a simple c-like type (int, >>>>>> bool, string, etc) >>>>>> >>>>>> Where the item itself has fields, it is an agglomerate type and it is >>>>>> descended recursively. >>>>>> >>>>>> My aim is to set each leaf programmatically. >>>>>> >>>>>> So as I traverse the arborescance, I am collecting the field >>>>>> definitions for the leafs. >>>>>> >>>>>> So later, I am addressing the variable again, but trying to set one >>>>>> of its leafs by the field definition I saved. Do I also have to save the >>>>>> descriptor for each leaf? >>>>>> >>>>>> Can what I want to do be done? >>>>>> >>>>>> >>>>>> On Tuesday, 22 June 2021 at 11:52:00 UTC-4 [email protected] wrote: >>>>>> >>>>>>> It looks to me like r->report points to a vafmsg.HealthReport but >>>>>>> the field descriptor refers to a field in another message >>>>>>> (vafmsg.HardwareComponent). >>>>>>> >>>>>>> On Tue, Jun 22, 2021 at 7:42 AM J G <[email protected]> wrote: >>>>>>> >>>>>>>> Hello Adam, >>>>>>>> >>>>>>>> OK, I understand, so I've tried this, but I get an error. >>>>>>>> >>>>>>>> void my_set_value( class healthreport * r, const char * >>>>>>>> defaultvalue, const google::protobuf::FieldDescriptor * descriptor ) { >>>>>>>> >>>>>>>> auto reflection = r->report->GetReflection(); >>>>>>>> >>>>>>>> switch( descriptor->type() ) { >>>>>>>> case google::protobuf::FieldDescriptor::TYPE_STRING: { >>>>>>>> >>>>>>>> printf( "REQUESTED TYPE: STRING" EOL ); >>>>>>>> std::string s = defaultvalue; >>>>>>>> reflection->SetString( r->report, descriptor, s ); >>>>>>>> >>>>>>>> } >>>>>>>> break; >>>>>>>> >>>>>>>> default: >>>>>>>> >>>>>>>> printf( "REQUESTED TYPE %d NOT HANDLED" EOL, >>>>>>>> descriptor->type() ); >>>>>>>> break; >>>>>>>> >>>>>>>> } >>>>>>>> >>>>>>>> } >>>>>>>> Running the above produces the following output: >>>>>>>> >>>>>>>> REQUESTED TYPE: STRING >>>>>>>> [libprotobuf FATAL >>>>>>>> /var/tmp/portage/dev-libs/protobuf-3.15.8/work/protobuf-3.15.8/src/google/protobuf/generated_message_reflection.cc:111] >>>>>>>> >>>>>>>> Protocol Buffer reflection usage error: >>>>>>>> Method : google::protobuf::Reflection::SetString >>>>>>>> Message type: vafmsg.HealthReport >>>>>>>> Field : vafmsg.HardwareComponent.hardware_interface >>>>>>>> Problem : Field does not match message type. >>>>>>>> terminate called after throwing an instance of >>>>>>>> 'google::protobuf::FatalException' >>>>>>>> what(): Protocol Buffer reflection usage error: >>>>>>>> Method : google::protobuf::Reflection::SetString >>>>>>>> Message type: vafmsg.HealthReport >>>>>>>> Field : vafmsg.HardwareComponent.hardware_interface >>>>>>>> Problem : Field does not match message type. >>>>>>>> >>>>>>>> Here is the proto definition of the variable triggering the >>>>>>>> exception: >>>>>>>> >>>>>>>> message HardwareComponent { >>>>>>>> optional Component component = 1; >>>>>>>> repeated DiscreteValue temp = 2; >>>>>>>> optional string hardware_interface = 3; >>>>>>>> optional uint32 remaining_life = 4; >>>>>>>> optional uint32 total_hours = 5; >>>>>>>> optional EnergyInfo energy = 6; >>>>>>>> } >>>>>>>> >>>>>>>> So the type really IS string, yet an exception is triggered... >>>>>>>> >>>>>>>> What am I doing wrong? >>>>>>>> >>>>>>>> On Friday, 18 June 2021 at 17:56:57 UTC-4 [email protected] wrote: >>>>>>>> >>>>>>>>> Each descriptor describes part of the schema (e.g. a message type, >>>>>>>>> enum type, etc.) but is unrelated to any particular instance of it. >>>>>>>>> As a >>>>>>>>> result, if you have a descriptor by itself then you can't really >>>>>>>>> modify >>>>>>>>> anything because you separately need an instance of the thing you >>>>>>>>> want to >>>>>>>>> modify. The way to programmatically modify a message is to use the >>>>>>>>> Reflection >>>>>>>>> <https://github.com/protocolbuffers/protobuf/blob/9d9d8ee18dedfb18371031cd299d1d282ddf707f/src/google/protobuf/message.h#L452> >>>>>>>>> >>>>>>>>> API. You can use Reflection::ListFields() to get a list of all the >>>>>>>>> fields >>>>>>>>> that are set on the message and then there are Reflection::Get* and >>>>>>>>> Reflection::Set* methods to get and set particular fields. >>>>>>>>> >>>>>>>>> On Fri, Jun 18, 2021 at 9:11 AM J G <[email protected]> wrote: >>>>>>>>> >>>>>>>>>> Hi, >>>>>>>>>> >>>>>>>>>> I've got some code that lets me recursively walk a protobuf >>>>>>>>>> variable. >>>>>>>>>> >>>>>>>>>> That part works, I can enumerate the characteristics of a >>>>>>>>>> variable, but the pointers/references returned by the API are const. >>>>>>>>>> >>>>>>>>>> My question is: From a variable's Descriptor or FieldDescriptor, >>>>>>>>>> is it possible to get a non-const pointer/reference to the field to >>>>>>>>>> be able >>>>>>>>>> to modify it? >>>>>>>>>> >>>>>>>>>> Here's my (simplified) code so far: >>>>>>>>>> >>>>>>>>>> void enumpb( const google::protobuf::Descriptor * d ) { >>>>>>>>>> >>>>>>>>>> for ( int i = 0; i < d->field_count(); i++ ) { >>>>>>>>>> >>>>>>>>>> auto field = d->field( i ); >>>>>>>>>> >>>>>>>>>> // Modify variable code here >>>>>>>>>> [...] >>>>>>>>>> >>>>>>>>>> auto mt = field->message_type(); >>>>>>>>>> if ( ! mt ) { >>>>>>>>>> continue; >>>>>>>>>> } else if ( 0 != strcmp( d->full_name().c_str(), >>>>>>>>>> mt->full_name().c_str() ) ) { >>>>>>>>>> >>>>>>>>>> enumpb( mt ) ; >>>>>>>>>> >>>>>>>>>> } >>>>>>>>>> >>>>>>>>>> } >>>>>>>>>> >>>>>>>>>> return true; >>>>>>>>>> >>>>>>>>>> } >>>>>>>>>> >>>>>>>>>> -- >>>>>>>>>> You received this message because you are subscribed to the >>>>>>>>>> Google Groups "Protocol Buffers" group. >>>>>>>>>> To unsubscribe from this group and stop receiving emails from it, >>>>>>>>>> send an email to [email protected]. >>>>>>>>>> To view this discussion on the web visit >>>>>>>>>> https://groups.google.com/d/msgid/protobuf/dcf6bb53-24ce-4404-ab71-0fe3a94adc40n%40googlegroups.com >>>>>>>>>> >>>>>>>>>> <https://groups.google.com/d/msgid/protobuf/dcf6bb53-24ce-4404-ab71-0fe3a94adc40n%40googlegroups.com?utm_medium=email&utm_source=footer> >>>>>>>>>> . >>>>>>>>>> >>>>>>>>> -- >>>>>>>> You received this message because you are subscribed to the Google >>>>>>>> Groups "Protocol Buffers" group. >>>>>>>> To unsubscribe from this group and stop receiving emails from it, >>>>>>>> send an email to [email protected]. >>>>>>>> >>>>>>> To view this discussion on the web visit >>>>>>>> https://groups.google.com/d/msgid/protobuf/7ef6e77e-8635-4b16-b570-b80f75d207d9n%40googlegroups.com >>>>>>>> >>>>>>>> <https://groups.google.com/d/msgid/protobuf/7ef6e77e-8635-4b16-b570-b80f75d207d9n%40googlegroups.com?utm_medium=email&utm_source=footer> >>>>>>>> . >>>>>>>> >>>>>>> -- >>>>>> You received this message because you are subscribed to the Google >>>>>> Groups "Protocol Buffers" group. >>>>>> To unsubscribe from this group and stop receiving emails from it, >>>>>> send an email to [email protected]. >>>>>> >>>>> To view this discussion on the web visit >>>>>> https://groups.google.com/d/msgid/protobuf/39fd7a89-426a-453d-9482-4cb3e3658da0n%40googlegroups.com >>>>>> >>>>>> <https://groups.google.com/d/msgid/protobuf/39fd7a89-426a-453d-9482-4cb3e3658da0n%40googlegroups.com?utm_medium=email&utm_source=footer> >>>>>> . >>>>>> >>>>> -- >>>> You received this message because you are subscribed to the Google >>>> Groups "Protocol Buffers" group. >>>> To unsubscribe from this group and stop receiving emails from it, send >>>> an email to [email protected]. >>>> >>> To view this discussion on the web visit >>>> https://groups.google.com/d/msgid/protobuf/764459e6-6f37-42aa-869c-0f8405aa13c6n%40googlegroups.com >>>> >>>> <https://groups.google.com/d/msgid/protobuf/764459e6-6f37-42aa-869c-0f8405aa13c6n%40googlegroups.com?utm_medium=email&utm_source=footer> >>>> . >>>> >>> -- >> You received this message because you are subscribed to the Google Groups >> "Protocol Buffers" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected]. >> > To view this discussion on the web visit >> https://groups.google.com/d/msgid/protobuf/3d9ed924-341b-43cf-8359-c91a93d1916dn%40googlegroups.com >> >> <https://groups.google.com/d/msgid/protobuf/3d9ed924-341b-43cf-8359-c91a93d1916dn%40googlegroups.com?utm_medium=email&utm_source=footer> >> . >> > -- You received this message because you are subscribed to the Google Groups "Protocol Buffers" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/protobuf/9c617cbc-f632-46a3-a1ac-c95c7fe6d2e2n%40googlegroups.com.
