[
https://issues.apache.org/jira/browse/THRIFT-5497?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17477465#comment-17477465
]
Richard H. Gumpertz edited comment on THRIFT-5497 at 1/17/22, 10:51 PM:
------------------------------------------------------------------------
I have three important questions:
(1) What should the *default ::apache::thrift::to_string* format be for
displaying binary_strings? Embedding lots of *\x* sequences is ugly and hard
to read: *"\Xde\Xad\Xbe\Xef"* !!!
I'm very much leaning toward {*}X"deadbeef"{*}. One could also consider the
C++ syntax *"deadbeef"_{_}X{_}* _(using syntax like a C++ _user-defined literal
operator_ ""_X, but that could conflict with an user application's use of the
operator ""_X). Furthermore, it has a readability problem: one's eye tends not
to see the X when it is at the end of the string; it it much more
attention-getting at the beginning. *_Anyway, PLEASE SUBMIT SUGGESTIONS!_*
(2) Due to unfamiliarity with the general layout of the Thrift include files, I
am a bit unsure where to put the definition of
:{*}:apache::thrift::binary_string{*}. I believe the header file Thrift.h is
the appropriate place but would very much appreciate confirmation. *{_}Any
feedback from the more experienced Thrift community{_}?*
(3) Is *::apache::thrift::to_string* used just for displaying strings to users
or is it also used in non-binary protocols for passing data? If the latter,
must any reader(s) or other functions be able to parse strings generated by
to_string? *_If so, which?_*
Finally, here is the latest working definition of
::apache::thrift::binary_string: [^binary_string_2022.01.17_#2.h] . I believe
it may finally be fairly stable,
was (Author: JIRAUSER283319):
I have three important questions:
(1) What should the *default ::apache::thrift::to_string* format be for display
blobs? embedding lots of *\x* sequences is ugly: *"\Xde\Xad\Xbe\Xef"* !!!
I'm very much leaning toward {*}X"deadbeef"{*}. One could also consider the
C++ syntax *"deadbeef"{_}X{_}* _(using syntax like a C++ _user-defined literal
operator_ ""_X, but that could conflict with an application's use of operator
""_X). Furthermore, it has a readability problem: one's eye tends not to see
the X at the end of the string; it it much more attention getting at the
beginning. PLEASE SUBMIT SUGGESTIONS!
(2) Due to unfamiliarity with the general layout of the Thrift include files, I
am a bit unsure where to put the definition of
:{*}:apache::thrift::binary_string{*}. I believe the header file Thrift.h is
the appropriate place but would very much appreciate confirmation. Any
feedback from the more experienced Thrift community?
(3) Is *::apache::thrift::to_string* used just for displaying strings to users
or is it also used in non-binary protocols for passing data? If the latter,
must any reader(s) or other functions be able to parse strings generated by
to_string? If so, which? **
Finally, here is the latest working definition of
::apache::thrift::binary_string: [^binary_string_2022.01.17_#2.h] . I believe
it may finally be fairly stable,
> Change the C++ representation of binary be std::basic_string<unsigned_char>
> ---------------------------------------------------------------------------
>
> Key: THRIFT-5497
> URL: https://issues.apache.org/jira/browse/THRIFT-5497
> Project: Thrift
> Issue Type: Improvement
> Reporter: Richard H. Gumpertz
> Priority: Minor
> Attachments: binary_string-1.h, binary_string.h,
> binary_string_2022.01.16_#2.h, binary_string_2022.01.17.h,
> binary_string_2022.01.17_#2.h
>
>
> I would like to change the C++ representation of binary from "std::string" to
> "std::basic_string<unsigned_char>". This would allow C++ templates to
> distinguish the Thrift type "string" from "binary".
> If nothing else, ::apached::thrift::to_string could print binary differently
> than string. Currently, it can get pretty ugly when one calls printTo on a
> struct containing a binary field that has data that is not readable!
> There is an issue as to how they should be displayed by default; using
> \xDE\xAD\xBE\xEF with a \ before every byte seems ugly. Alternativlely,
> "DEADBEEF"_x because it could also be used as input if the C++ operator""_x
> is defined by the user? Alternatively just <binary> or <27-byte binary>
> (mirroring <null> for unset fields), hiding the contents? Perhaps leave
> to_string(std::basic_string<unsigned char> &value) undefined and force the
> user to define it if needed? Thoughts & suggestions???
> Given that such a change could break existing code, it should probably be
> enabled under a new thrift compiler option to --gen cpp. I haven't come up
> with a good name for such an option yet: --gen cpp:binary_as_unsigned_string
> is explicit but wordy. Any suggestions???
> I would be willing to do the work necessary to add the relevant code if there
> are no volunteers who have a better understanding of precisely what code
> would have to be updated (or at least help me find such code)
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
