Fix per-table file identifiers and extensions

[olivier-matz-6wind]

(Extracted from PR #185)

These 2 commits change how per-table file identifiers and extensions are generated:

• the file identifiers should not use flatbuffers_identifier because it prevents from including several schema definitions, use a direct string instead
• the per-table file extensions are not generated, let's add them similarly to file identifiers

changes in v2:

• fix documentation (README.md)

[mikkelfj]

There is something strange with the dot prefix in the file extension. "mon" is not listed as ".mon" in monster_test.fbs

In the log of the test build file_extension is found with ".cgen_test" in one place, and "cgen_test" without dot in some other place.

[mikkelfj]

A reference output is printed without dot: https://github.com/dvidelabs/flatcc/blob/2eaf0fa3a18e5fcab7fc66f04accb677119f2e27/test/cgen_test/cgen_test.c#L110

I wonder both why the test does not fail against the reference, and also why the dot is placed in the generated output.

[mikkelfj]

The dot is generated here, in existing code:

https://github.com/dvidelabs/flatcc/blob/2eaf0fa3a18e5fcab7fc66f04accb677119f2e27/src/compiler/codegen_c_reader.c#L1012

I'm not sure that is correct to have a dot prefix, but it would be a breaking change to remove it.

[olivier-matz-6wind]

The dot is generated here, in existing code:

https://github.com/dvidelabs/flatcc/blob/2eaf0fa3a18e5fcab7fc66f04accb677119f2e27/src/compiler/codegen_c_reader.c#L1012

I'm not sure that is correct to have a dot prefix, but it would be a breaking change to remove it.

Yes, you are right, I'll add the dot in next version.

Note: in case we use the default bin (out->opts->default_bin_ext), the dot is already present, so no change needed for this code branch.

[mikkelfj]

I think it would be reasonable to make a breaking change here and remove the dot from the output.

Also note that the prerecorded output uses "file_extension" where the generated output uses flatbuffers_extension. I guess there is no diff executing in the text.

[mikkelfj]

Note: in case we use the default bin (out->opts->default_bin_ext), the dot is already present, so no change needed for this code branch.

Yes the default is defined with a dot, but the question is whether it should be. I think it should, but the otherwise the extension should be as define in the schema unmodified. Not really sure here.

[olivier-matz-6wind]

Without the dot, it looks more consistent with the schema. I can try to remove the dot everywhere then (will do in another commit, before this one).

[mikkelfj]

good

[olivier-matz-6wind]

changes in v3:

• add a new commit to remove the dot from generated file extension (i.e. #define flatbuffers_extension)
• rework last commit to remove the dot too

I chose to keep FLATCC_DEFAULT_BIN_EXT and opts->default_bin_ext with the dot, for consistency with other extension macros and variables. The dot is removed when printed, using (opts->default_bin_ext + 1). Let me know if you are ok with this.

[mikkelfj]

+1 is not good without checking for length and dot.

I would have suggested to change the config file, but as you say, other ext vars use a dot prefix. There isn't really a good answer here. I would prefer if the dot was added in the schema and the default var kept as is, but that is not convention.

I think the best option is to remove dot from config var and add a comment that schema file identifiers do not carry a dot by convention. That way users can also define their own ext if they want.

[olivier-matz-6wind]

+1 is not good without checking for length and dot.

Oh yes, that's right

I would have suggested to change the config file, but as you say, other ext vars use a dot prefix. There isn't really a good answer here. I would prefer if the dot was added in the schema and the default var kept as is, but that is not convention.

I think the best option is to remove dot from config var and add a comment that schema file identifiers do not carry a dot by convention. That way users can also define their own ext if they want.

Ok, will do, thanks

[olivier-matz-6wind]

changes in v4:

• remove dot from FLATCC_DEFAULT_BIN_EXT config macro to avoid the unsafe (opts->default_bin_ext + 1)

[olivier-matz-6wind]

changes in v5:

• Fix comment style

[mikkelfj]

thanks

[olivier-matz-6wind]

thanks for the quick and efficient review!