Initial Bonn Integration

[matthewspear]

Integrating Bonn RGB-D Dynamic Dataset into SLAMBench. Added as WIP to allow better viablity.

Will submit PR after tidying up the new approach to regex!

[matthewspear]

Need some feedback on a design approach:

I am tidying up the way regex is used across Bonn

Method A:

const std::string& t = RegexPattern::timestamp;
const std::string& w = RegexPattern::whitespace;
const std::string& n = RegexPattern::number;
const std::string& start = RegexPattern::start;
const std::string& end = RegexPattern::end;

// format: timestamp tx ty tz qx qy qz qw
std::vector<std::string> expression{start,
                                    t, w,       // timestamp
                                    n, w,       // tx
                                    n, w,       // ty
                                    n, w,       // tz
                                    n, w,       // qx
                                    n, w,       // qy
                                    n, w,       // qz
                                    n, end};    // qw

boost::regex expr = boost::regex(std::accumulate(expression.begin(), expression.end(), std::string()));

Instead of vector + accumulate these could be combined value the + operator.

Method B:

// format: timestamp tx ty tz qx qy qz qw
std::vector<std::string> expression2{RegexPattern::start,
                                        RegexPattern::timestamp, RegexPattern::whitespace, // timestamp
                                        RegexPattern::number, RegexPattern::whitespace,    // tx
                                        RegexPattern::number, RegexPattern::whitespace,    // ty
                                        RegexPattern::number, RegexPattern::whitespace,    // tz
                                        RegexPattern::number, RegexPattern::whitespace,    // qx
                                        RegexPattern::number, RegexPattern::whitespace,    // qy
                                        RegexPattern::number, RegexPattern::whitespace,    // qz
                                        RegexPattern::number, RegexPattern::end};          // qw

boost::regex expr = boost::regex(std::accumulate(expression.begin(), expression.end(), std::string()));

The main difference is having a copy for clarity and avoiding repeating RegexPattern too many times.

@mihaibujanca what do you think?

[matthewspear]

Slightly simpler:

const std::string& t = RegexPattern::timestamp;
const std::string& w = RegexPattern::whitespace;
const std::string& n = RegexPattern::number;
const std::string& start = RegexPattern::start;
const std::string& end = RegexPattern::end;

// format: timestamp tx ty tz qx qy qz qw
std::string expr = start
                  + t + w       // timestamp
                  + n + w       // tx
                  + n + w       // ty
                  + n + w       // tz
                  + n + w       // qx
                  + n + w       // qy
                  + n + w       // qz
                  + n + end;    // qw

boost::regex groundtruth = boost::regex(expr);

[mihaibujanca]

Slightly simpler:

const std::string& t = RegexPattern::timestamp;
const std::string& w = RegexPattern::whitespace;
const std::string& n = RegexPattern::number;
const std::string& start = RegexPattern::start;
const std::string& end = RegexPattern::end;

// format: timestamp tx ty tz qx qy qz qw
std::string expr = start
                  + t + w       // timestamp
                  + n + w       // tx
                  + n + w       // ty
                  + n + w       // tz
                  + n + w       // qx
                  + n + w       // qy
                  + n + w       // qz
                  + n + end;    // qw

boost::regex groundtruth = boost::regex(expr);

I quite like this. Perhaps also a midway between clarity and simplicity is using "time" or "ts" instead of "t" and "num" instead of "n"?

[matthewspear]

Cool I'll go for that - saves the unnecessary use of vector and accumulate, and I agree about the names!!

[mihaibujanca]

Good job!

Perhaps one suggestion would be moving RegexPattern.h to something like dataset-tools/include/utils/RegexPattern.h.

It might seem a bit overkill when there's just one file in the directory but we don't know what else we might add there in the future (e.g perhaps the loadDatasetSensor kind of functions?)

[matthewspear]

Perhaps one suggestion would be moving RegexPattern.h to something like dataset-tools/include/utils/RegexPattern.h.

Yeah sure, that makes sense – I'm working on the refactoring / other utils in another branch called dataset-tools that'll do a draft PR for soon!

[mihaibujanca]

Happy for this to be merged whenever you're done with it, nicely done!