PROJ RFC 2: Initial integration of "GDAL SRS barn" work
- Author:
Even Rouault
- Contact:
even.rouault at spatialys.com
- Status:
Adopted, implemented in PROJ 6.0
- Initial version:
2018-10-09
- Last Updated:
2018-10-31
Summary
This RFC is the result of a first phase of the GDAL Coordinate System Barn Raising efforts. In its current state, this work mostly consists of:
a C++ implementation of the ISO-19111:2018 / OGC Topic 2 "Referencing by coordinates" classes to represent Datums, Coordinate systems, CRSs (Coordinate Reference Systems) and Coordinate Operations.
methods to convert between this C++ modeling and WKT1, WKT2 and PROJ string representations of those objects
management and query of a SQLite3 database of CRS and Coordinate Operation definition
a C API binding part of those capabilities
Details
Structure in packages / namespaces
The C++ implementation of the (upcoming) ISO-19111:2018 / OGC Topic 2 "Referencing by coordinates" classes follows this abstract modeling as much as possible, using package names as C++ namespaces, abstract classes and method names. A new BoundCRS class has been added to cover the modeling of the WKT2 BoundCRS construct, that is a generalization of the WKT1 TOWGS84 concept. It is strongly recommended to have the ISO-19111 standard open to have an introduction for the concepts when looking at the code. A few classes have also been inspired by the GeoAPI
The classes are organized into several namespaces:
- osgeo::proj::util
A set of base types from ISO 19103, GeoAPI and other PROJ "technical" specific classes
Template optional<T>, classes BaseObject, IComparable, BoxedValue, ArrayOfBaseObject, PropertyMap, LocalName, NameSpace, GenericName, NameFactory, CodeList, Exception, InvalidValueTypeException, UnsupportedOperationException
- osgeo::proj::metadata:
Common classes from ISO 19115 (Metadata) standard
Classes Citation, GeographicExtent, GeographicBoundingBox, TemporalExtent, VerticalExtent, Extent, Identifier, PositionalAccuracy,
- osgeo::proj::common:
Common classes: UnitOfMeasure, Measure, Scale, Angle, Length, DateTime, DateEpoch, IdentifiedObject, ObjectDomain, ObjectUsage
- osgeo::proj::cs:
Coordinate systems and their axis
Classes AxisDirection, Meridian, CoordinateSystemAxis, CoordinateSystem, SphericalCS, EllipsoidalCS, VerticalCS, CartesianCS, OrdinalCS, ParametricCS, TemporalCS, DateTimeTemporalCS, TemporalCountCS, TemporalMeasureCS
- osgeo::proj::datum:
Datum (the relationship of a coordinate system to the body)
Classes Ellipsoid, PrimeMeridian, Datum, DatumEnsemble, GeodeticReferenceFrame, DynamicGeodeticReferenceFrame, VerticalReferenceFrame, DynamicVerticalReferenceFrame, TemporalDatum, EngineeringDatum, ParametricDatum
- osgeo::proj::crs:
CRS = coordinate reference system = coordinate system with a datum
Classes CRS, GeodeticCRS, GeographicCRS, DerivedCRS, ProjectedCRS, VerticalCRS, CompoundCRS, BoundCRS, TemporalCRS, EngineeringCRS, ParametricCRS, DerivedGeodeticCRS, DerivedGeographicCRS, DerivedProjectedCRS, DerivedVerticalCRS
- osgeo::proj::operation:
Coordinate operations (relationship between any two coordinate reference systems)
Classes CoordinateOperation, GeneralOperationParameter, OperationParameter, GeneralParameterValue, ParameterValue, OperationParameterValue, OperationMethod, InvalidOperation, SingleOperation, Conversion, Transformation, PointMotionOperation, ConcatenatedOperation
- osgeo::proj::io:
I/O classes: WKTFormatter, PROJStringFormatter, FormattingException, ParsingException, IWKTExportable, IPROJStringExportable, WKTNode, WKTParser, PROJStringParser, DatabaseContext, AuthorityFactory, FactoryException, NoSuchAuthorityCodeException
What does what?
The code to parse WKT and PROJ strings and build ISO-19111 objects is contained in io.cpp
The code to format WKT and PROJ strings from ISO-19111 objects is mostly contained in the related exportToWKT() and exportToPROJString() methods overridden in the applicable classes. io.cpp contains the general mechanics to build such strings.
Regarding WKT strings, three variants are handled in import and export:
WKT2_2018: variant corresponding to the upcoming ISO-19162:2018 standard
WKT2_2015: variant corresponding to the current ISO-19162:2015 standard
WKT1_GDAL: variant corresponding to the way GDAL understands the OGC 01-099 and OGC 99-049 standards
Regarding PROJ strings, two variants are handled in import and export:
PROJ5: variant used by PROJ >= 5, possibly using pipeline constructs, and avoiding +towgs84 / +nadgrids legacy constructs. This variant honours axis order and input/output units. That is the pipeline for the conversion of EPSG:4326 to EPSG:32631 will assume that the input coordinates are in latitude, longitude order, with degrees.
PROJ4: variant used by PROJ 4.x
The raw query of the proj.db database and the upper level construction of ISO-19111 objects from the database contents is done in factory.cpp
A few design principles
Methods generally take and return xxxNNPtr objects, that is non-null shared pointers (pointers with internal reference counting). The advantage of this approach is that the user has not to care about the life-cycle of the instances (and this makes the code leak-free by design). The only point of attention is to make sure no reference cycles are made. This is the case for all classes, except the CoordinateOperation class that point to CRS for sourceCRS and targetCRS members, whereas DerivedCRS point to a Conversion instance (which derives from CoordinateOperation). This issue was detected in the ISO-19111 standard. The solution adopted here is to use std::weak_ptr in the CoordinateOperation class to avoid the cycle. This design artifact is transparent to users.
Another important design point is that all ISO19111 objects are immutable after creation, that is they only have getters that do not modify their states. Consequently they could possibly use in a thread-safe way. There are however classes like PROJStringFormatter, WKTFormatter, DatabaseContext, AuthorityFactory and CoordinateOperationContext whose instances are mutable and thus can not be used by multiple threads at once.
Example how to build the EPSG:4326 / WGS84 Geographic2D definition from scratch:
auto greenwich = PrimeMeridian::create(
util::PropertyMap()
.set(metadata::Identifier::CODESPACE_KEY,
metadata::Identifier::EPSG)
.set(metadata::Identifier::CODE_KEY, 8901)
.set(common::IdentifiedObject::NAME_KEY, "Greenwich"),
common::Angle(0));
// actually predefined as PrimeMeridian::GREENWICH constant
auto ellipsoid = Ellipsoid::createFlattenedSphere(
util::PropertyMap()
.set(metadata::Identifier::CODESPACE_KEY, metadata::Identifier::EPSG)
.set(metadata::Identifier::CODE_KEY, 7030)
.set(common::IdentifiedObject::NAME_KEY, "WGS 84"),
common::Length(6378137),
common::Scale(298.257223563));
// actually predefined as Ellipsoid::WGS84 constant
auto datum = GeodeticReferenceFrame::create(
util::PropertyMap()
.set(metadata::Identifier::CODESPACE_KEY, metadata::Identifier::EPSG)
.set(metadata::Identifier::CODE_KEY, 6326)
.set(common::IdentifiedObject::NAME_KEY, "World Geodetic System 1984");
ellipsoid
util::optional<std::string>(), // anchor
greenwich);
// actually predefined as GeodeticReferenceFrame::EPSG_6326 constant
auto geogCRS = GeographicCRS::create(
util::PropertyMap()
.set(metadata::Identifier::CODESPACE_KEY, metadata::Identifier::EPSG)
.set(metadata::Identifier::CODE_KEY, 4326)
.set(common::IdentifiedObject::NAME_KEY, "WGS 84"),
datum,
cs::EllipsoidalCS::createLatitudeLongitude(scommon::UnitOfMeasure::DEGREE));
// actually predefined as GeographicCRS::EPSG_4326 constant
Algorithmic focus
On the algorithmic side, a somewhat involved logic is the CoordinateOperationFactory::createOperations() in coordinateoperation.cpp that takes a pair of source and target CRS and returns a set of possible coordinate operations (either single operations like a Conversion or a Transformation, or concatenated operations). It uses the intrinsic structure of those objects to create the coordinate operation pipeline. That is, if going from a ProjectedCRS to another one, by doing first the inverse conversion from the source ProjectedCRS to its base GeographicCRS, then finding the appropriate transformation(s) from this base GeographicCRS to the base GeographicCRS of the target CRS, and then applying the conversion from this base GeographicCRS to the target ProjectedCRS. At each step, it queries the database to find if one or several transformations are available. The resulting coordinate operations are filtered, and sorted, with user provided hints:
desired accuracy
area of use, defined as a bounding box in longitude, latitude space (its actual CRS does not matter for the intended use)
if no area of use is defined, if and how the area of use of the source and target CRS should be used. By default, the smallest area of use is used. The rationale is for example when transforming between a national ProjectedCRS and a world-scope GeographicCRS to use the are of use of this ProjectedCRS to select the appropriate datum shifts.
how the area of use of the candidate transformations and the desired area of use (either explicitly or implicitly defined, as explained above) are compared. By default, only transformations whose area of use is fully contained in the desired area of use are selected. It is also possible to relax this test by specifying that only an intersection test must be used.
whether PROJ transformation grid names should be substituted to the official names, when a match is found in the grid_alternatives table of the database. Defaults to true
whether the availability of those grids should be used to filter and sort the results. By default, the transformations using grids available in the system will be presented first.
The results are sorted, with the most relevant ones appearing first in the result vector. The criteria used are in that order
grid actual availability: operations referencing grids not available will be listed after ones with available grids
grid potential availability: operation referencing grids not known at all in the proj.db will be listed after operations with grids known, but not available.
known accuracy: operations with unknown accuracies will be listed after operations with known accuracy
area of use: operations with smaller area of use (the intersection of the operation area of used with the desired area of use) will be listed after the ones with larger area of use
accuracy: operations with lower accuracy will be listed after operations with higher accuracy (caution: lower accuracy actually means a higher numeric value of the accuracy property, since it is a precision in metre)
All those settings can be specified in the CoordinateOperationContext instance passed to createOperations().
An interesting example to understand how those parameters play together is to use projinfo -s EPSG:4267 -t EPSG:4326 (NAD27 to WGS84 conversions), and see how specifying desired area of use, spatial criterion, grid availability, etc. affects the results.
The following command currently returns 78 results:
projinfo -s EPSG:4267 -t EPSG:4326 --summary --spatial-test intersects
The createOperations() algorithm also does a kind of "CRS routing". A typical example is if wanting to transform between CRS A and CRS B, but no direct transformation is referenced in proj.db between those. But if there are transformations between A <--> C and B <--> C, then it is possible to build a concatenated operation A --> C --> B. The typical example is when C is WGS84, but the implementation is generic and just finds a common pivot from the database. An example of finding a non-WGS84 pivot is when searching a transformation between EPSG:4326 and EPSG:6668 (JGD2011 - Japanese Geodetic Datum 2011), which has no direct transformation registered in the EPSG database . However there are transformations between those two CRS and JGD2000 (and also Tokyo datum, but that one involves less accurate transformations)
projinfo -s EPSG:4326 -t EPSG:6668 --grid-check none --bbox 135.42,34.84,142.14,41.58 --summary
Candidate operations found: 7
unknown id, Inverse of JGD2000 to WGS 84 (1) + JGD2000 to JGD2011 (1), 1.2 m, Japan - northern Honshu
unknown id, Inverse of JGD2000 to WGS 84 (1) + JGD2000 to JGD2011 (2), 2 m, Japan excluding northern main province
unknown id, Inverse of Tokyo to WGS 84 (108) + Tokyo to JGD2011 (2), 9.2 m, Japan onshore excluding northern main province
unknown id, Inverse of Tokyo to WGS 84 (108) + Tokyo to JGD2000 (2) + JGD2000 to JGD2011 (1), 9.4 m, Japan - northern Honshu
unknown id, Inverse of Tokyo to WGS 84 (2) + Tokyo to JGD2011 (2), 13.2 m, Japan - onshore mainland and adjacent islands
unknown id, Inverse of Tokyo to WGS 84 (2) + Tokyo to JGD2000 (2) + JGD2000 to JGD2011 (1), 13.4 m, Japan - northern Honshu
unknown id, Inverse of Tokyo to WGS 84 (1) + Tokyo to JGD2011 (2), 29.2 m, Asia - Japan and South Korea
Code repository
The current state of the work can be found in the iso19111 branch of rouault/proj.4 repository , and is also available as a GitHub pull request at https://github.com/OSGeo/proj.4/pull/1040
Here is a not-so-usable comparison with a fixed snapshot of master branch
Database
Content
The database contains CRS and coordinate operation definitions from the EPSG database (IOGP’s EPSG Geodetic Parameter Dataset) v9.5.3, IGNF registry (French National Geographic Institute), ESRI database, as well as a few customizations.
Building (for PROJ developers creating the database)
The building of the database is a several stage process:
Construct SQL scripts for EPSG
The first stage consists in constructing .sql scripts mostly with CREATE TABLE and INSERT statements to create the database structure and populate it. There is one .sql file for each database table, populated with the content of the EPSG database, automatically generated with the build_db.py script, which processes the PostgreSQL dumps issued by IOGP. A number of other scripts are dedicated to manual editing, for example grid_alternatives.sql file that binds official grid names to PROJ grid names
Concert UTF8 SQL to sqlite3 db
The second stage is done automatically by the make process. It pipes the .sql script, in the right order, to the sqlite3 binary to generate a first version of the proj.db SQLite3 database.
Add extra registries
The third stage consists in creating additional .sql files from the content of other registries. For that process, we need to bind some definitions of those registries to those of the EPSG database, to be able to link to existing objects and detect some boring duplicates. The ignf.sql file has been generated using the build_db_create_ignf.py script from the current data/IGNF file that contains CRS definitions (and implicit transformations to WGS84) as PROJ.4 strings. The esri.sql file has been generated using the build_db_from_esri.py script, from the .csv files in https://github.com/Esri/projection-engine-db-doc/tree/master/csv
Finalize proj.db
The last stage runs make again to incorporate the new .sql files generated in the previous stage (so the process of building the database involves a kind of bootstrapping...)
Building (for PROJ users)
The make process just runs the second stage mentioned above from the .sql files. The resulting proj.db is currently 5.3 MB large.
Structure
The database is structured into the following tables and views. They generally match a ISO-19111 concept, and is generally close to the general structure of the EPSG database. Regarding identification of objects, where the EPSG database only contains a 'code' numeric column, the PROJ database identifies objects with a (auth_name, code) tuple of string values, allowing several registries to be combined together.
- Technical:
authority_list: view enumerating the authorities present in the database. Currently: EPSG, IGNF, PROJ
metadata: a few key/value pairs, for example to indicate the version of the registries imported in the database
object_view: synthetic view listing objects (ellipsoids, datums, CRS, coordinate operations...) code and name, and the table name where they are further described
alias_names: list possible alias for the name field of object table
link_from_deprecated_to_non_deprecated: to handle the link between old ESRI to new ESRI/EPSG codes
- Common:
unit_of_measure: table with UnitOfMeasure definitions.
area: table with area-of-use (bounding boxes) applicable to CRS and coordinate operations.
- Coordinate systems:
axis: table with CoordinateSystemAxis definitions.
coordinate_system: table with CoordinateSystem definitions.
- Ellipsoid and datums:
ellipsoid: table with ellipsoid definitions.
prime_meridian: table with PrimeMeridian definitions.
geodetic_datum: table with GeodeticReferenceFrame definitions.
vertical_datum: table with VerticalReferenceFrame definitions.
- CRS:
geodetic_crs: table with GeodeticCRS and GeographicCRS definitions.
projected_crs: table with ProjectedCRS definitions.
vertical_crs: table with VerticalCRS definitions.
compound_crs: table with CompoundCRS definitions.
- Coordinate operations:
coordinate_operation_view: view giving a number of common attributes shared by the concrete tables implementing CoordinateOperation
conversion: table with definitions of Conversion (mostly parameter and values of Projection)
concatenated_operation: table with definitions of ConcatenatedOperation.
grid_transformation: table with all grid-based transformations.
grid_packages: table listing packages in which grids can be found. ie "proj-datumgrid", "proj-datumgrid-europe", ...
grid_alternatives: table binding official grid names to PROJ grid names. e.g "Und_min2.5x2.5_egm2008_isw=82_WGS84_TideFree.gz" --> "egm08_25.gtx"
helmert_transformation: table with all Helmert-based transformations.
other_transformation: table with other type of transformations.
The main departure with the structure of the EPSG database is the split of the various coordinate operations over several tables. This was done mostly for human-readability as the EPSG organization of coordoperation, coordoperationmethod, coordoperationparam, coordoperationparamusage, coordoperationparamvalue tables makes it hard to grasp at once all the parameters and values for a given operation.
Utilities
A new projinfo utility has been added. It enables the user to enter a CRS or coordinate operation by a AUTHORITY:CODE, PROJ string or WKT string, and see it translated in the different flavors of PROJ and WKT strings. It also enables to build coordinate operations between two CRSs.
Usage
usage: projinfo [-o formats] [-k crs|operation] [--summary] [-q]
[--bbox min_long,min_lat,max_long,max_lat]
[--spatial-test contains|intersects]
[--crs-extent-use none|both|intersection|smallest]
[--grid-check none|discard_missing|sort]
[--boundcrs-to-wgs84]
{object_definition} | (-s {srs_def} -t {srs_def})
-o: formats is a comma separated combination of: all,default,PROJ4,PROJ,WKT_ALL,WKT2_2015,WKT2_2018,WKT1_GDAL
Except 'all' and 'default', other format can be preceded by '-' to disable them
Examples
Specify CRS by PROJ string and specify output formats
$ projinfo -o PROJ4,PROJ,WKT1_GDAL,WKT2_2018 "+title=IGN 1972 Nuku Hiva - UTM fuseau 7 Sud +proj=tmerc +towgs84=165.7320,216.7200,180.5050,-0.6434,-0.4512,-0.0791,7.420400 +a=6378388.0000 +rf=297.0000000000000 +lat_0=0.000000000 +lon_0=-141.000000000 +k_0=0.99960000 +x_0=500000.000 +y_0=10000000.000 +units=m +no_defs"
PROJ string:
Error when exporting to PROJ string: BoundCRS cannot be exported as a PROJ.5 string, but its baseCRS might
PROJ.4 string:
+proj=utm +zone=7 +south +ellps=intl +towgs84=165.732,216.72,180.505,-0.6434,-0.4512,-0.0791,7.4204
WKT2_2018 string:
BOUNDCRS[
SOURCECRS[
PROJCRS["IGN 1972 Nuku Hiva - UTM fuseau 7 Sud",
BASEGEOGCRS["unknown",
DATUM["unknown",
ELLIPSOID["International 1909 (Hayford)",6378388,297,
LENGTHUNIT["metre",1,
ID["EPSG",9001]]]],
PRIMEM["Greenwich",0,
ANGLEUNIT["degree",0.0174532925199433],
ID["EPSG",8901]]],
CONVERSION["unknown",
METHOD["Transverse Mercator",
ID["EPSG",9807]],
PARAMETER["Latitude of natural origin",0,
ANGLEUNIT["degree",0.0174532925199433],
ID["EPSG",8801]],
PARAMETER["Longitude of natural origin",-141,
ANGLEUNIT["degree",0.0174532925199433],
ID["EPSG",8802]],
PARAMETER["Scale factor at natural origin",0.9996,
SCALEUNIT["unity",1],
ID["EPSG",8805]],
PARAMETER["False easting",500000,
LENGTHUNIT["metre",1],
ID["EPSG",8806]],
PARAMETER["False northing",10000000,
LENGTHUNIT["metre",1],
ID["EPSG",8807]]],
CS[Cartesian,2],
AXIS["(E)",east,
ORDER[1],
LENGTHUNIT["metre",1,
ID["EPSG",9001]]],
AXIS["(N)",north,
ORDER[2],
LENGTHUNIT["metre",1,
ID["EPSG",9001]]]]],
TARGETCRS[
GEOGCRS["WGS 84",
DATUM["World Geodetic System 1984",
ELLIPSOID["WGS 84",6378137,298.257223563,
LENGTHUNIT["metre",1]]],
PRIMEM["Greenwich",0,
ANGLEUNIT["degree",0.0174532925199433]],
CS[ellipsoidal,2],
AXIS["latitude",north,
ORDER[1],
ANGLEUNIT["degree",0.0174532925199433]],
AXIS["longitude",east,
ORDER[2],
ANGLEUNIT["degree",0.0174532925199433]],
ID["EPSG",4326]]],
ABRIDGEDTRANSFORMATION["Transformation from unknown to WGS84",
METHOD["Position Vector transformation (geog2D domain)",
ID["EPSG",9606]],
PARAMETER["X-axis translation",165.732,
ID["EPSG",8605]],
PARAMETER["Y-axis translation",216.72,
ID["EPSG",8606]],
PARAMETER["Z-axis translation",180.505,
ID["EPSG",8607]],
PARAMETER["X-axis rotation",-0.6434,
ID["EPSG",8608]],
PARAMETER["Y-axis rotation",-0.4512,
ID["EPSG",8609]],
PARAMETER["Z-axis rotation",-0.0791,
ID["EPSG",8610]],
PARAMETER["Scale difference",1.0000074204,
ID["EPSG",8611]]]]
WKT1_GDAL:
PROJCS["IGN 1972 Nuku Hiva - UTM fuseau 7 Sud",
GEOGCS["unknown",
DATUM["unknown",
SPHEROID["International 1909 (Hayford)",6378388,297],
TOWGS84[165.732,216.72,180.505,-0.6434,-0.4512,-0.0791,7.4204]],
PRIMEM["Greenwich",0,
AUTHORITY["EPSG","8901"]],
UNIT["degree",0.0174532925199433,
AUTHORITY["EPSG","9122"]],
AXIS["Longitude",EAST],
AXIS["Latitude",NORTH]],
PROJECTION["Transverse_Mercator"],
PARAMETER["latitude_of_origin",0],
PARAMETER["central_meridian",-141],
PARAMETER["scale_factor",0.9996],
PARAMETER["false_easting",500000],
PARAMETER["false_northing",10000000],
UNIT["metre",1,
AUTHORITY["EPSG","9001"]],
AXIS["Easting",EAST],
AXIS["Northing",NORTH]]
Find transformations between 2 CRS
Between "Poland zone I" (based on Pulkovo 42 datum) and "UTM WGS84 zone 34N"
Summary view:
$ projinfo -s EPSG:2171 -t EPSG:32634 --summary
Candidate operations found: 1
unknown id, Inverse of Poland zone I + Pulkovo 1942(58) to WGS 84 (1) + UTM zone 34N, 1 m, Poland - onshore
Display of pipelines:
$ PROJ_LIB=data src/projinfo -s EPSG:2171 -t EPSG:32634 -o PROJ
PROJ string:
+proj=pipeline +step +proj=axisswap +order=2,1 +step +inv +proj=sterea +lat_0=50.625 +lon_0=21.0833333333333 +k=0.9998 +x_0=4637000 +y_0=5647000 +ellps=krass +step +proj=cart +ellps=krass +step +proj=helmert +x=33.4 +y=-146.6 +z=-76.3 +rx=-0.359 +ry=-0.053 +rz=0.844 +s=-0.84 +convention=position_vector +step +inv +proj=cart +ellps=WGS84 +step +proj=utm +zone=34 +ellps=WGS84
Impacted files
New files (excluding makefile.am, CMakeLists.txt and other build infrastructure artifacts):
- include/proj/: Public installed C++ headers
common.hpp: declarations of osgeo::proj::common namespace.
coordinateoperation.hpp: declarations of osgeo::proj::operation namespace.
coordinatesystem.hpp: declarations of osgeo::proj::cs namespace.
crs.hpp: declarations of osgeo::proj::crs namespace.
datum.hpp: declarations of osgeo::proj::datum namespace.
io.hpp: declarations of osgeo::proj::io namespace.
metadata.hpp: declarations of osgeo::proj::metadata namespace.
util.hpp: declarations of osgeo::proj::util namespace.
nn.hpp: Code from https://github.com/dropbox/nn to manage Non-nullable pointers for C++
- include/proj/internal: Private non-installed C++ headers
coordinateoperation_internal.hpp: classes InverseCoordinateOperation, InverseConversion, InverseTransformation, PROJBasedOperation, and functions to get conversion mappings between WKT and PROJ syntax
coordinateoperation_constants.hpp: Select subset of conversion/transformation EPSG names and codes for the purpose of translating them to PROJ strings
coordinatesystem_internal.hpp: classes AxisDirectionWKT1, AxisName and AxisAbbreviation
internal.hpp: a few helper functions, mostly to do string-based operations
io_internal.hpp: class WKTConstants
helmert_constants.hpp: Helmert-based transformation & parameters names and codes.
lru_cache.hpp: code from https://github.com/mohaps/lrucache11 to have a generic Least-Recently-Used cache of objects
- src/:
c_api.cpp: C++ API mapped to C functions
common.cpp: implementation of common.hpp
coordinateoperation.cpp: implementation of coordinateoperation.hpp
coordinatesystem.cpp: implementation of coordinatesystem.hpp
factory.cpp: implementation of AuthorityFactory class (from io.hpp)
internal.cpp: implementation of internal.hpp
metadata.cpp: implementation of metadata.hpp
static.cpp: a number of static constants (like pre-defined well-known ellipsoid, datum and CRS), put in the right order for correct static initializations
projinfo.cpp: new 'projinfo' binary
general.dox: generic introduction documentation.
- data/sql/:
area.sql: generated by build_db.py
axis.sql: generated by build_db.py
begin.sql: hand generated (trivial)
commit.sql: hand generated (trivial)
compound_crs.sql: generated by build_db.py
concatenated_operation.sql: generated by build_db.py
conversion.sql: generated by build_db.py
coordinate_operation.sql: generated by build_db.py
coordinate_system.sql: generated by build_db.py
crs.sql: generated by build_db.py
customizations.sql: hand generated (empty)
ellipsoid.sql: generated by build_db.py
geodetic_crs.sql: generated by build_db.py
geodetic_datum.sql: generated by build_db.py
grid_alternatives.sql: hand-generated. Contains links between official registry grid names and PROJ ones
grid_transformation.sql: generated by build_db.py
grid_transformation_custom.sql: hand-generated
helmert_transformation.sql: generated by build_db.py
ignf.sql: generated by build_db_create_ignf.py
esri.sql: generated by build_db_from_esri.py
metadata.sql: hand-generated
other_transformation.sql: generated by build_db.py
prime_meridian.sql: generated by build_db.py
proj_db_table_defs.sql: hand-generated. Database structure: CREATE TABLE / CREATE VIEW / CREATE TRIGGER
projected_crs.sql: generated by build_db.py
unit_of_measure.sql: generated by build_db.py
vertical_crs.sql: generated by build_db.py
vertical_datum.sql: generated by build_db.py
- scripts/:
build_db.py : generate .sql files from EPSG database dumps
build_db_create_ignf.py: generates data/sql/ignf.sql
build_db_from_esri.py: generates data/sql/esri.sql
doxygen.sh: generates Doxygen documentation
gen_html_coverage.sh: generates HTML report of the coverage for --coverage build
filter_lcov_info.py: utility used by gen_html_coverage.sh
reformat.sh: used by reformat_cpp.sh
reformat_cpp.sh: reformat all .cpp/.hpp files according to LLVM-style formatting rules
- tests/unit/
test_c_api.cpp: test of src/c_api.cpp
test_common.cpp: test of src/common.cpp
test_util.cpp: test of src/util.cpp
test_crs.cpp: test of src/crs.cpp
test_datum.cpp: test of src/datum.cpp
test_factory.cpp: test of src/factory.cpp
test_io.cpp: test of src/io.cpp
test_metadata.cpp: test of src/metadata.cpp
test_operation.cpp: test of src/operation.cpp
C API
proj.h has been extended to bind a number of C++ classes/methods to a C API.
The main structure is an opaque PJ_OBJ* roughly encapsulating a osgeo::proj::BaseObject, that can represent a CRS or a CoordinateOperation object. A number of the C functions will work only if the right type of underlying C++ object is used with them. Misuse will be properly handled at runtime. If a user passes a PJ_OBJ* representing a coordinate operation to a pj_obj_crs_xxxx() function, it will properly error out. This design has been chosen over creating a dedicate PJ_xxx object for each C++ class, because such an approach would require adding many conversion and free functions for little benefit.
This C API is incomplete. In particular, it does not allow to build ISO19111 objects at hand. However it currently permits a number of actions:
building CRS and coordinate operations from WKT and PROJ strings, or from the proj.db database
exporting CRS and coordinate operations as WKT and PROJ strings
querying main attributes of those objects
finding coordinate operations between two CRS.
test_c_api.cpp should demonstrates simple usage of the API (note: for the conveniency of writing the tests in C++, test_c_api.cpp wraps the C PJ_OBJ* instances in C++ 'keeper' objects that automatically call the pj_obj_unref() function at function end. In a pure C use, the caller must use pj_obj_unref() to prevent leaks.)
Documentation
All public C++ classes and methods and C functions are documented with Doxygen.
Current snapshot of Class list
A basic integration of the Doxygen XML output into the general PROJ documentation (using reStructuredText format) has been done with the Sphinx extension Breathe, producing:
Testing
Nearly all exported methods are tested by a unit test. Global line coverage of the new files is 92%. Those tests represent 16k lines of codes.
Build requirements
The new code leverages on a number of C++11 features (auto keyword, constexpr, initializer list, std::shared_ptr, lambda functions, etc.), which means that a C++11-compliant compiler must be used to generate PROJ:
gcc >= 4.8
clang >= 3.3
Visual Studio >= 2015.
Compilers tested by the Travis-CI and AppVeyor continuous integration environments:
GCC 4.8
mingw-w64-x86-64 4.8
clang 5.0
Apple LLVM version 9.1.0 (clang-902.0.39.2)
MSVC 2015 32 and 64 bit
MSVC 2017 32 and 64 bit
The libsqlite3 >= 3.7 development package must also be available. And the sqlite3 binary must be available to build the proj.db files from the .sql files.
Runtime requirements
libc++/libstdc++/MSVC runtime consistent with the compiler used
libsqlite3 >= 3.7
Backward compatibility
At this stage, no backward compatibility issue is foreseen, as no existing functional C code has been modified to use the new capabilities
Future work
The work described in this RFC will be pursued in a number of directions. Non-exhaustively:
Support for ESRI WKT1 dialect (PROJ currently ingest the ProjectedCRS in esri.sql in that dialect, but there is no mapping between it and EPSG operation and parameter names, so conversion to PROJ strings does not always work.
closer integration with the existing code base. In particular, the +init=dict:code syntax should now go first to the database (then the epsg and IGNF files can be removed). Similarly proj_create_crs_to_crs() could use the new capabilities to find an appropriate coordinate transformation.
and whatever else changes are needed to address GDAL and libgeotiff needs
Adoption status
The RFC has been adopted with support from PSC members Kurt Schwehr, Kristian Evers, Howard Butler and Even Rouault.