c++ - Using boost parse datetime string: With single digit hour format

Question

I am working on code which needs to compile on NDK toolchain. Unfortunately, latest version only supports till gcc4.9 which does not support C++11 datetime parsing. I have a date time string which I need to send thru two-three formats to figure out parsing method.

So I tried linux API stftime which sometimes give values on wrong parsing method. I had to abandon it and move to boost.

Now coming on to boost I am using 1_64 version. According to the documentation here

I could not find a way to parse single digit hour format.

bool getepochtime(const std::string &str, const std::string &format, unsigned long &epoch){
    epoch = 0;
    namespace bt = boost::posix_time;
    std::locale lformat = std::locale(std::locale::classic(), new bt::time_input_facet(format));
    bt::ptime pt;
    std::istringstream is(str);
    is.imbue(lformat);
    is >> pt;
    if (pt == bt::ptime()) {
        //epoch = 0;
        return false;
    }
    bt::ptime timet_start(boost::gregorian::date(1970, 1, 1));
    bt::time_duration diff = pt - timet_start;
    epoch = (1000 * diff.ticks()/bt::time_duration::rep_type::ticks_per_second);
    return true;
}

int main() {    
    unsigned long eval;
    // this works.
    getepochtime("28th january 11:50 PM", "%dth %B %H:%M %p", eval);
    // this does not work.
    getepochtime("28th january 1:50 PM", "%dth %B %I:%M %p", eval);
    // nor this.
    getepochtime("28th january 1:50 PM", "%dth %B %H:%M %p", eval);
    return 0;
}

Any help will be appreciated.

Answer 1

I'll leave it to you to sort out how you want dates without years to be interpreted. However, here's a quick start using /just/ strptime.

I used it in a larger codebase, and we needed some pretty versatile date recognition. Behold: the adaptive datetime parser:

#pragma once

#include <string>
#include <chrono>
#include <cstdint>
#include <list>

namespace mylib { namespace datetime { 

/*
 * Multi-format capable date time parser
 *
 * Intended to be seeded with a list of supported formats, in order of
 * preference. By default, parser is not adaptive (mode is `fixed`).
 *
 * In adaptive modes the format can be required to be 
 *
 *  - sticky (consistently reuse the first matched format)
 *  - ban_failed (remove failed patterns from the list; banning only occurs
 *    on successful parse to avoid banning all patterns on invalid input)
 *  - mru (preserves the list but re-orders for performance)
 *
 * CAUTION: 
 *   If formats are ambiguous (e.g. mm-dd-yyyy vs dd-mm-yyyy) allowing
 *   re-ordering results in unpredictable results.
 *   => Only use `mru` when there are no ambiguous formats
 *
 * NOTE: 
 *   The function object is stateful. In algorithms, pass it by reference
 *   (`std::ref(obj)`) to avoid copying the patterns and to ensure correct
 *   adaptive behaviour
 *
 * NOTE:
 *   - use %z before %Z to correctly handle [-+]hhmm POSIX TZ indications
 *   - adaptive_parser is thread-safe as long as it's not in any adaptive
 *     mode (the only allowed flag is `full_match`)
 */
 class adaptive_parser {
   public:
     typedef std::list<std::string> list_t;

     enum mode_t {
         fixed      = 0, // not adapting; keep trying same formats in same order
         sticky     = 1, // re-use first successful format consistently
         ban_failed = 2, // forget formats that have failed
         mru        = 4, // optimize by putting last known good in front
         full_match = 8, // require full matches to be accepted
     };

     adaptive_parser(mode_t m = full_match);
     adaptive_parser(mode_t m, list_t formats);

     // returns seconds since epoch
     std::chrono::seconds operator()(std::string);

   private:
     mode_t _mode;
     list_t _formats;
};

static inline adaptive_parser::mode_t operator|(adaptive_parser::mode_t lhs, adaptive_parser::mode_t rhs) {
    return static_cast<adaptive_parser::mode_t>(static_cast<int>(lhs) | static_cast<int>(rhs)); 
}

} }

You can use it as such:

Live On Wandbox

#include "adaptive_parser.h"
#include <string>
#include <iostream>

int main() {    
    using namespace mylib::datetime;

    adaptive_parser parser { adaptive_parser::full_match, {
            "%Y %dth %B %H:%M %p",
               "%dth %B %H:%M %p",
            "%Y %dth %B %I:%M %p",
               "%dth %B %I:%M %p",
        } };

    for (std::string const input : {
            "2017 28th january 11:50 PM",
            "28th january 11:50 PM",
            "2017 28th january 1:50 PM",
            "28th january 1:50 PM",
        })
    try {
        std::cout << "Parsing '" << input << "'
";
        std::cout << " -> epoch " << parser(input).count() << "
";
    } catch(std::exception const& e) {
        std::cout << "Exception: " << e.what() << "
";
    }
}

Printing:

Parsing '2017 28th january 11:50 PM'
 -> epoch 1485604200
Parsing '28th january 11:50 PM'
 -> epoch -2206613400
Parsing '2017 28th january 1:50 PM'
 -> epoch 1485568200
Parsing '28th january 1:50 PM'
 -> epoch -2206649400

Note that epoch -2206613400 corresponds to 28 jan 1900

Implementation

The implementation comes with a bunch of pretty well-tuned unambiguous date patterns. Our project used a number of "hacks" to normalize strange input formats, these have been omitted (you can see the commented references to detail::normalize_... functions for ideas):

#include "adaptive_parser.h"
#include "time.h"
#include <vector>
#include <algorithm>
#include <cassert>
#include <cstring>
#include <iostream>

namespace {
    enum level { LOG_DEBUG };
    static std::ostream s_devnull { nullptr };

    struct {
        std::ostream& log(int) const {
#ifdef NDEBUG
            return s_devnull;
#else
            return std::cerr;
#endif
        };
    } s_trace;
}

namespace mylib { namespace datetime { 

    adaptive_parser::adaptive_parser(mode_t m) 
        : _mode(m), _formats {
// use EOL_MARK to debug patterns when you suspect ambiguity or partial matches
#define EOL_MARK "" // " EOL_MARK"
// use %z before %Z to correctly handle [-+]hhmm POSIX time zone offsets
#if __GLIBC__ == 2 && __GLIBC_MINOR__ <= 15
    // ubuntu 12.04 used eglibc and doesn't parse all bells and whistles
#define WITH_TZ(prefix, suffix) prefix " %z" suffix, prefix " %Z" suffix, prefix " Z" suffix, prefix " (UTC)" suffix, prefix suffix
#else
#define WITH_TZ(prefix, suffix) prefix " %z" suffix, prefix " %Z" suffix, prefix suffix
#endif
            WITH_TZ("%Y-%m-%dT%H:%M:%S.%f", EOL_MARK),
            WITH_TZ("%Y-%m-%dT%H:%M:%S", EOL_MARK),
            WITH_TZ("%Y-%m-%dT%H:%M", EOL_MARK),
            // 
            WITH_TZ("%Y-%m-%dT%I:%M:%S.%f %p", EOL_MARK),
            WITH_TZ("%Y-%m-%dT%I:%M:%S %p", EOL_MARK),
            WITH_TZ("%Y-%m-%dT%I:%M %p", EOL_MARK),
            // 
            WITH_TZ("%Y-%m-%d%n%H:%M:%S", EOL_MARK),
            WITH_TZ("%Y-%m-%d%n%I:%M:%S %p", EOL_MARK),
            //
            WITH_TZ("%a %b %d %H:%M:%S %Y", EOL_MARK),
            WITH_TZ("%a %b %d %I:%M:%S %p %Y", EOL_MARK),
            //
            WITH_TZ("%a %d %b %H:%M:%S %Y", EOL_MARK),
            WITH_TZ("%a %d %b %I:%M:%S %p %Y", EOL_MARK),
            //
            WITH_TZ("%a, %b %d %H:%M:%S %Y", EOL_MARK),
            WITH_TZ("%a, %b %d %I:%M:%S %p %Y", EOL_MARK),
            //
            WITH_TZ("%a, %d %b %H:%M:%S %Y", EOL_MARK),
            WITH_TZ("%a, %d %b %I:%M:%S %p %Y", EOL_MARK),
            //////
            WITH_TZ("%a %d %b %Y %H:%M:%S", EOL_MARK),
            WITH_TZ("%a %d %b %Y %I:%M:%S %p", EOL_MARK),
            //
            WITH_TZ("%a, %d %b %Y %H:%M:%S", EOL_MARK),
            WITH_TZ("%a, %d %b %Y %I:%M:%S %p", EOL_MARK),
#undef WITH_TZ
            /*
             * HUMAN DATE:
             *
             * This pattern would ambiguate the "%s" one (sadly, because it
             * leads to obviously bogus results like parsing "1110871987" into
             * "2063-04-24 16:25:59" (because "1110-8-7T19:8:7" matches
             * "%Y-%m-%dT%H:%M:%S %Z" somehow...).
             *
             * We work around this issue by normalizing detected
             * 'yyyyMMddhhmmss' human dates into iso format as a preprocessing
             * step.
             */
            //"%Y %m %d %H %M %S" EOL_MARK,

            // epoch seconds
            "@%s" EOL_MARK,
            "%s" EOL_MARK,
           }
    { }

    adaptive_parser::adaptive_parser(mode_t m, list_t formats)
        : _mode(m), _formats(std::move(formats))
    { }

    std::chrono::seconds adaptive_parser::operator()(std::string input) {
        if (_formats.empty()) throw std::invalid_argument("No candidate patterns in datetime::adaptive_parser");
        if (input.empty()) throw std::invalid_argument("Empty input cannot be parsed as a date time");

        //detail::normalize_tz(input);
        //detail::normalize_tz_utc_w_offset_re(input);
        //detail::normalize_date_sep(input);
        //detail::normalize_human_date(input);
        //detail::normalize_redundant_timezone_description(input);
        input += EOL_MARK;

        std::vector<list_t::iterator> failed;

        bool matched = false;
        struct tm time_struct;

        auto pattern = _formats.begin();
        for (; !matched && pattern != _formats.end(); ++pattern) {
            memset(&time_struct, 0, sizeof(time_struct));
            auto tail = ::strptime(input.c_str(), pattern->c_str(), &time_struct);

            matched = tail;
            //if (matched) s_trace.log(LOG_DEBUG) << "Input '" << input << "' successfully matched pattern '" << *pattern << "' leaving '" << tail << "'
";

            if (_mode & full_match) {
                while (tail && *tail && std::isspace(*tail))
                    ++tail; // skip trailing whitespace
                matched &= tail && !*tail;
            }

            if (matched)
                break;

            if (_mode & ban_failed)
                failed.push_back(pattern);
        }

        if (matched) {
            for (auto to_ban : failed) {
                s_trace.log(LOG_DEBUG) << "Banning failed datetime pattern: " << *to_ban << "
";
                _formats.erase(to_ban);
            }

            if (_mode & sticky) {
                s_trace.log(LOG_DEBUG) << "Made succeeding datetime pattern sticky: " << *pattern << "
";
                _formats = { *pattern };
            }

            if ((_mode & mru) && pattern != _formats.begin()) {
                assert(pattern != _formats.end()); // inconsistent with `matched==true`

                s_trace.log(LOG_DEBUG) << "Promote succeeding datetime pattern to the top: " << *pattern << "
";
                std::rotate(_formats.begin(), pattern, std::next(pattern));
            }
#ifdef __FreeBSD__
            auto raw = (time_struct.tm_gmtoff)? mktime(&time_struct) : timegm(&time_struct);
            return std::chrono::seconds(raw);
#else
            long offset = time_struct.tm_gmtoff;
            return std::chrono::seconds(timegm (&time_struct) - offset);
#endif
        }

        s_trace.log(LOG_DEBUG) << "Failed to parse datetime input '" << input << "' with " << _formats.size() << " patterns
";
        throw std::runtime_error("Input cannot be parsed as a date time");
    }

} }