Serialization Formal Review #2
From the library synopsis: Here, we use the term "serialization" to mean the reversible deconstruction of an arbitrary set of C++ data structures to a sequence of bytes. Such a system can be used to reconstitute an equivalent structure in another program context. Depending on this context, this might used implement object
All - Today (April 13th) is the start of the formal review of the Serialization library by Robert Ramey. The review will run until Monday April 26th. I will be serving as review manager. Library details are below. I have a number of special requests for this review: 1) This is a large, complex, and important library. Please don't wait till the end of the review period to start your review -- there is no additional room in the review schedule to extend the review period. 2) I am particularly interested in hearing from parties that have tried to extend the library by adding serialization for a new type or by building a new type of archive. The ability to easily extend the library for new types and novel types of archives is critical for this library. 3) If you reviewed the library during the first review, I would like your evaluation of whether the new version of the library addresses your original concerns. Also, if this is a re-review I would appreciate if you could send me a weblink from the mail archive to your original review posting(s). 4) Finally, I would like reviewers to indicate parts of the library that might become standalone boost libraries. While this will have little bearing on the acceptance / rejection of the library it might allow a reduction in the size of the library. As usual, please state in review comments how you reviewed the library and whether the you think the library should be accepted into Boost. Further guidelines for writing reviews can be found on the website at: http://www.boost.org/more/formal_review_process.htm#Comments ********************************************** Library Details: ********************************************** persistence, remote parameter passing or other facility. In this system we use the term "archive" to refer to a specific rendering of this stream of bytes. This could be a file of binary data, text data, XML, or some other created by the user of this library. The latest package can be found at the following locations: http://www.rrsd.com/boost/index.htm http://groups.yahoo.com/group/boost/files/serialization18.zip Thanks, Jeff
Hi Robert, After a few tweaks - which I will report later - I got serialization to compile on my installation of MacOS X. Before submitting my review I'm looking trough the documentation to see how I can implement an optimized archive along the lines of demo_fast_binary_archive.cpp, but extended to also support std::vector, std::valarray, ... At this point I have a question that does not seem to be addressed in the documentation: What is the difference betweem save and save_override or load and load_override? demo_fast_binary_archive.cpp uses save_override and load_override but I could not find it documented? It seems to me that the new serialization library allows me to do what I want but I'm confused regarding these functions. Best regards Matthias
Hi Robert, This is some comments based on reading the tutorial. I think your explanations are fine, but I would like to see some remarks added: eg class gps_position { public: int degrees; int minutes; float seconds; gps_position(){}; gps_position(int d, int m, float s) : degrees(d), minutes(m), seconds(s) {} }; I think it should be stated that this is not recommended practice, ie, this class has a non-trivial invariant and should not have public members. A better example might be to use std::pair. The examples use pointers in arrays and pointers in standard containers. I think the text should state that this is not recommended; or even better, throw in a shared_ptr instead. The code in the example is inherently exception unsafe. Therefore I also think that container< shared_ptr<T> > should be supported by default whereas container< T* > is, well, ok if you don't have exceptions somehow. br Thorsten
This is a reminder that the Serialization review is more than half over with ~5 days remaining. Thx for your time. Jeff On Mon, 12 Apr 2004 22:09:36 -0700, Jeff Garland wrote
All -
Today (April 13th) is the start of the formal review of the Serialization library by Robert Ramey. The review will run until Monday April 26th. I will be serving as review manager. Library details are below. I have a number of special requests for this review:
1) This is a large, complex, and important library. Please don't wait till the end of the review period to start your review -- there is no additional room in the review schedule to extend the review period. 2) I am particularly interested in hearing from parties that have tried to extend the library by adding serialization for a new type or by building a new type of archive. The ability to easily extend the library for new types and novel types of archives is critical for this library. 3) If you reviewed the library during the first review, I would like your evaluation of whether the new version of the library addresses your original concerns. Also, if this is a re- review I would appreciate if you could send me a weblink from the mail archive to your original review posting(s). 4) Finally, I would like reviewers to indicate parts of the library that might become standalone boost libraries. While this will have little bearing on the acceptance / rejection of the library it might allow a reduction in the size of the library.
As usual, please state in review comments how you reviewed the library and whether the you think the library should be accepted into Boost. Further guidelines for writing reviews can be found on the website at:
http://www.boost.org/more/formal_review_process.htm#Comments
********************************************** Library Details: ********************************************** From the library synopsis: Here, we use the term "serialization" to mean the reversible deconstruction of an arbitrary set of C++ data structures to a sequence of bytes. Such a system can be used to reconstitute an equivalent structure in another program context. Depending on this context, this might used implement object persistence, remote parameter passing or other facility. In this system we use the term "archive" to refer to a specific rendering of this stream of bytes. This could be a file of binary data, text data, XML, or some other created by the user of this library.
The latest package can be found at the following locations:
http://www.rrsd.com/boost/index.htm http://groups.yahoo.com/group/boost/files/serialization18.zip
Thanks,
Jeff
On Apr 12, 2004, at 10:09 PM, Jeff Garland wrote:
2) I am particularly interested in hearing from parties that have tried to extend the library by adding serialization for a new type or by building a new type of archive. The ability to easily extend the library for new types and novel types of archives is critical for this library.
I have tried extending the library since I want to swap our old serialization library used at http://alps.comp-phys.org with a Boost serialization library as soon as possible. It was important to keep the same archive format, since we have gigabytes of data collected over ten years in millions of CPU-hours and need to be able to read those with both new and old codes. With the help of Robert I managed to convince myself that this is indeed possible. The only issue I have is that the documentation needs to be improved.
3) If you reviewed the library during the first review, I would like your evaluation of whether the new version of the library addresses your original concerns. Also, if this is a re-review I would appreciate if you could send me a weblink from the mail archive to your original review posting(s).
In the first round I voted against inclusion in Boost because i) the design was not flexible enough to allow reading my old legacy formats and ii) the design inhibted some optimizations I need for large data sets. Robert did an excellent job in addressing this and other issues. The new version is greatly improved and I am satisfied with the design.
As usual, please state in review comments how you reviewed the library and whether the you think the library should be accepted into Boost. Further guidelines for writing reviews can be found on the website at:
http://www.boost.org/more/formal_review_process.htm#Comments
What is your evaluation of the design?
The current design offers the flexibility needed for all serialization tasks I can think of and makes it possible to implement serialization without any performance penalty.
What is your evaluation of the implementation?
I did not look at many details of the implementation but have a few comments: 1. there are still a few "dangerous" code portions, such as load_binary(const_cast<char *>(s.data()), l); in line 89 of file basic_binary_iprimitive.ipp . It is stated in a comment that this might be unsafe, but it is probably OK on all current implementations of the standard. 2. the comments in the source files are at several places out of sync with the implementation, they sometimes still refer to old designed of the library. A thorough check of all files is needed. 3. looking at the implementation after reading the documentation is confusing at some places because of tricks used to work around compiler deficiencies. For example, the documentation discusses load() and save() functions but the implementations use load_override() and save_override() functions. I would prefer to have the implementation be consistent with the documentation and the workarounds included only in #if .... #endif guards.
What is your evaluation of the documentation?
The documentation is OK for using the library, but more and better documentation is needed for: 1. implementing new archive types: there, as Robert is aware, the example in the documentation is incorrect. A more extended documentation would be very useful. 2. Better documentation for the classes used to track classes, class version, objects, etc.. For example, in newarchive.html there is a table listing some types (class_id_optional_type is listed twice???) but I am missing a) more information about the purpose of these types b) I found some additional types such as class_name_type, which is used in the implementation of some archives but not documented. 3. Documentation for the workarounds used to placate deficient compilers (e.g. save_override, ...) is missing 4. Documentation of archive formats, especially what class and object information is stored. I want to be able to predict, from reading the documentation, what exactly will be written to the archive and in which order. This is essential when exchanging data with an application that does not use this library.
What is your evaluation of the potential usefulness of the library?
In my opinion this will be among the most useful and most widely Boost libraries.
Did you try to use the library? With what compiler? Did you have any problems?
I tried using the library on MacOS X 10.3 using Apple's version of the g++ 3.3 compiler. I encountered two problems. 1. archive /wcslen.hpp does not compile because for some unknown reason boost/config.hpp defines BOOST_NO_CWCHAR, although the <cwchar> header exists. Is there a deep reason for this configuration option on MacOS X or is this a bug in the configuration? 2. trying to compile text_wiarchive.cpp and xml_wiarchive.cpp I get error messages such as boost/archive/iterators/remove_whitespace.hpp:146: error: ` boost::archive::iterators::filter_iterator<Predicate, Base>::m_predicate' has incomplete type I did not investigate this further at this time because of lack of time. and because I do not currently need the wide stream archive types.
How much effort did you put into your evaluation? A glance? A quick reading? In-depth study?
Being busy and traveling from Switzerland to California has not allowed me time for an in-depth study of all aspects. I focused compiling the library under MacOS X, writing short example programs to studdy the archive formats with, and implementing a new archive type.
Are you knowledgeable about the problem domain?
I have use serialization in a portable binary format on a daily basis for a decade. To summarize, I vote for accepting the library into Boost but would like to see the documentation improved. Matthias
Here's my review: ***** What is your evaluation of the design? ***** The overall design seems sound. However, apart from what other reviewers mentioned I have a few additional suggestions and nitpicks (I did not have the time to follow all the discussions, so some points may have been brough up already): 1. On compilers supporting ADL, it seems the user-supplied serialization function for non-intrusive serialization template<class Archive> void serialize(Archive & ar, gps_position & g, const unsigned int version); could also be put into the namespace of the type that it serializes. If this is correct it should be mentioned in the tutorial & reference. If not, there should be a rationale why not. 2. save/load naming: All functions that have to do with reading from an archive are named load*, all functions that have to do with writing to an archive are named save*. To me (non-native speaker) these names are generally associated with persistence only. Since the library is more general (one could use it for network communication, etc.) load/save are a bit missleading. I'd rather see read/write or other names that more clearly communicate what the functions are doing. 3. archive naming: The word archive also seems to be associated with persistence. I think serializer would much better describe the functionality. Another option I could live with is serialization_stream (I slightly disagree with the assertion in archives.html that an archive as modelled by this library is not a stream. It might not be what C++ folks traditionally understand a stream is but to me it definitely is a more general stream, i.e. a stream of objects/things/whatever). 4. Object tracking 1: Seeming contradiction I've had difficulty determining how exactly object tracking works. I believe the associated docs are slightly contradictory: - Under exceptions.html#pointer_conflict I read that the pointer_conflict exception is thrown when we save a class member through a pointer first and then "normally" (i.e. through a reference). - traits.html#tracking says "... That is[,] addresses of serialized objects are tracked if and only if an object of the same type is anywhere in the program serialized through a pointer." The second sentence seems to contradict the first one in that it does not specify that the sequence of saving is important (I assume that pointer_conflict exception is not thrown if we save normally first and then through a pointer). Moreover, I don't see how the library could possibly achieve the "anywhere in the program" bit. 5. Object tracking 2: It should not be possible to serialize pointers referencing objects of types that have track_never tracking type. special.html#objecttracking says: <quote> By definition, data types designated primitive by Implementation Level class serialization trait are never tracked. If it is desired to track a shared primitive object through a pointer (e.g. a long used as a reference count), It should be wrapped in a class/struct so that it is an identifiable type. The alternative of changing the implementation level of a long would affect all longs serialized in the whole program - probably not what one would intend. </quote> I think serializing untracked pointers almost never makes sense. If it does anyway the pointed-to object can always be saved by dereferencing the pointer. Such an error should probably be signalled with an exception. 6. Object tracking 3: One should be able to activate tracking on the fly Rarely it makes sense to save even primitive types in tracked mode (e.g. the example quoted in 6.). That's why I think setting the tracking type per type is not flexible enough. It should be possible to override the default somehow. I could imagine to save such pointers as follows: class T { ... int i; int * pI; // always points to i }; template<class Archive > void T::save(Archive &ar) const { ar << track( i ); // save the int here, keep track of the pointer ar << track( pI ); // only save a reference to the int } Since it is an error to save an untracked type by pointer even the second line needs the track( ) call. 7. I think the enum tracking_type should rather be named tracking_mode as the _type suffix suggests something else than an enum. 8. I don't understand how function instantiation as described under special.html#instantiation works. The description seems to indicate that all possible archive classes need to be included in the files of all types that are exported. Is that true? 9. The library registers all kinds of stuff with static instances. This precludes its use outside main(). This and other limitations should be stated explicitly in the rationale. ***** What is your evaluation of the implementation? ***** I did not have a look at the implementation. ***** What is your evaluation of the documentation? ***** 10. The tutorial is good. A few more links to advanced stuff could make it even easier to find your way around the library, e.g. how to split free serialization functions. 11. As other reviewers have already pointed out, the reference needs some serious work (this is not to say that Robert did a bad job, it's just not there yet). Some stuff is explained in "tutorial-language", i.e. too informal what left me wondering about the details (I can give examples if necessary). Moreover, the layout should be in C++ standard-style and include all the concepts, classes, functions and macros provided. ***** What is your evaluation of the potential usefulness of the library? ***** Very useful. ***** Did you try to use the library? With what compiler? Did you have any problems? ***** No, I did not try to use the library. ***** How much effort did you put into your evaluation? A glance? A quick reading? In-depth study? ***** About 6 hours reading the docs and writing this review. ***** Are you knowledgeable about the problem domain? ***** Sort of. 5 years ago I tried to write a library covering a small subset of what this library does. I eventually abadoned it because of compiler problems and lack of time. ***** Do you think the library should be accepted as a Boost library? Be sure to say this explicitly so that your other comments don't obscure your overall opinion. ***** Yes. Given that the points mentioned above are resolved I vote to accept the library into boost. A big thanks to Robert for all this excellent work! Andreas
Hi This is my re-review of the Serialization library. My previous review can be found here: http://lists.boost.org/MailArchives/boost/msg39608.php Most of the issues I previously had have been adressed in this version. ---------------------------------------------------------------------------- ------- * What is your evaluation of the design? Very good. It gives great flexibility to the user while still being very efficient. Compared to the previous version up for review this has become even better. At that time I wasn't so concerned by archive formats but has since come to appreciate the large improvements in that area. The new synax & interface for serialization is also very nice. * What is your evaluation of the implementation? As I mentioned in my previous review I would like to see it handle most of the STL and core boost components "out of the box". (see my attachment also) In particular I think the smart_ptrs (including auto_ptr) should have core-support from the serializer, even special cased if necessary (similar to boost::bind etc). The code to serialize std::auto_ptr could easily be copied to the official directory for example. Regarding shared_ & weak_ptr. If the implementation path Robert outlines in the documentation is to be used I guess an official patch to shared_count is necessary. Would be interesting to hear what the smart pointer experts think. Otherwise, I still think the solution I outlined in my previous review that doesn't require modifying shared_ptr could be considered. Downside beeing that it most likely adds a special-cased codepath to the serializer, but as I said I think it would be worth it. The code is nicely written altough complex. Highly nested template code and lots of mpl. All compile-time checks and further factorizations are valuable. Nitpick: serialization/scoped_ptr.hpp seems to have a typo. An extra #pragma once is present outside the #ifdef (_MSC..) block. It also includes serialization/pfto.hpp but doesn't seem to use it. * Parts of the library that might become standalone boost libraries The library already provides good separation of components. static_warning, strong typedef, pfto, stack_allocate etc are all given candidates that would just need better documentation and (fast-track?) reviews. * What is your evaluation of the documentation? I would also like to see an improved reference documentation that clearly states the requirements on the types/concepts involved. I support the separation of documentation (and implementation) of other boost components that like to support serialization. A simple index with links would suffice I think but most importantly a _consistent_ structure should be agreed upon which new components can follow. Nitpick: bibliography link to CommonC++ serializer is broken; should be: http://www.gnu.org/software/commoncpp/ * What is your evaluation of the potential usefulness of the library? Very useful! I'm looking forward to using the library in many projects to come. * Did you try to use the library? With what compiler? Yes. I used VC7.1. I compiled the review version with a current boost CVS and the only thing I had to change was to #ifdef the duplicate boost::is_abstract code. I would appreciate VC7.1 project files for the examples and tests though. One of my tests added support for the STL container adaptors std::stack, queue and priority_queue. I think direct support for them could be convinient since they hide the native containers. I attach my implementation. Feel free to comment. (The implementation could be considered to employ premature optimization.. ;-) ) * How much effort did you put into your evaluation? * Are you knowledgeable about the problem domain? Moderate. I've used the MFC serializer, developed and used an in-house "MFC copy", and used the CommonC++ Persist lib to which I added STL and boost::smart_ptr support. I've been trying out several of the previous versions of Robert's library ever since he first announced it on the boost-list almost two years ago. * Do you think the library should be accepted as a Boost library? Yes! Thanks for a fine library! // Fredrik Blomqvist begin 666 queue.hpp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` end begin 666 stack.hpp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`` ` end
participants (2)
-
Andreas Huber -
Fredrik Blomqvist -
Jeff Garland -
Matthias Troyer -
Thorsten Ottosen