Frequently asked questions

This page lists some Frequently Asked Questions (FAQ) when using Apache SIS.


Getting started

How do I transform a coordinate?

The following Java code projects a geographic coordinate from the World Geodetic System 1984 (WGS84) to WGS 84 / UTM zone 33N. In order to make the example a little bit simpler, this code uses predefined constants given by the CommonCRS convenience class. But more advanced applications will typically use EPSG codes instead. Note that all geographic coordinates below express latitude before longitude.

import org.opengis.geometry.DirectPosition;
import org.opengis.referencing.operation.CoordinateOperation;
import org.opengis.referencing.operation.TransformException;
import org.opengis.util.FactoryException;
import org.apache.sis.referencing.CRS;
import org.apache.sis.referencing.CommonCRS;
import org.apache.sis.geometry.DirectPosition2D;

public class MyApp {
    public static void main(String[] args) throws FactoryException, TransformException {
        CoordinateReferenceSystem sourceCRS = CommonCRS.WGS84.geographic();
        CoordinateReferenceSystem targetCRS = CommonCRS.WGS84.universal(40, 14);  // Get whatever zone is valid for 14°E.
        CoordinateOperation operation = CRS.findOperation(sourceCRS, targetCRS, null);
         * The above lines are costly and should be performed only once before to project many points.
         * In this example, the operation that we got is valid for coordinates in geographic area from
         * 12°E to 18°E (UTM zone 33) and 0°N to 84°N.
        DirectPosition ptSrc = new DirectPosition2D(40, 14);           // 40°N 14°E
        DirectPosition ptDst = operation.getMathTransform().transform(ptSrc, null);

        System.out.println("Source: " + ptSrc);
        System.out.println("Target: " + ptDst);

Which map projections are supported?

The operation methods (including, but not limited to, map projections) supported by Apache SIS are listed in the Coordinate Operation Methods page. The amount of map projection methods is relatively small, but the amount of projected Coordinate Reference Systems that we can build from them can be very large. For example with only three family of methods (Cylindrical Mercator, Transverse Mercator and Lambert Conic Conformal) used with different parameter values, we can cover thousands of projected CRS listed in the EPSG geodetic dataset.

In order to use a map projection method, we need to know the value to assign to the projection parameters. For convenience, thousands of projected CRS with pre-defined parameter values are are assigned a unique identifier. A well-known source of such definitions is the EPSG geodetic dataset, but other authorities also exist. The predefined CRS known to Apache SIS are listed in the Coordinate Reference Systems page.

What is the axis order issue and how is it addressed?

The axis order is specified by the authority (typically a national agency) defining the Coordinate Reference System (CRS). The order depends on the CRS type and the country defining the CRS. In the case of geographic CRS, the (latitude, longitude) axis order is widely used by geographers and pilots for centuries. However software developers tend to consistently use the (x, y) order for every kind of CRS. Those different practices resulted in contradictory definitions of axis order for almost every CRS of kind GeographicCRS, for some ProjectedCRS in the South hemisphere (South Africa, Australia, etc.) and for some polar projections among others.

For any CRS identified by an EPSG code, the official axis order can be checked on the official EPSG registry at (not to be confused with other sites having “epsg” in their name, but actually unrelated to the organization in charge of EPSG definitions): click on the “Retrieve by code” link and enter the numerical code. Then click on the “View” link on the right side, and click on the "+" symbol of the left side of “Axes”.

Recent OGC standards mandate the use of axis order as defined by the authority. Oldest OGC standards used the (x, y) axis order instead, ignoring any authority specification. Among the legacy OGC standards that used the non-conform axis order, an influent one is version 1 of the Well Known Text (WKT) format specification. According that widely-used format, WKT definitions without explicit AXIS[…] elements shall default to (longitude, latitude) or (x, y) axis order. In version 2 of the WKT format, AXIS[…] elements are no longer optional and should contain an explicit ORDER[…] sub-element for making the intended order yet more obvious.

Many software products still use the old (x, y) axis order, sometime because it is easier to implement. But Apache SIS rather defaults to axis order as defined by the authority (except when parsing a WKT 1 definition), but allows changing axis order to the (x, y) order after CRS creation. This change can be done with the following code:

CoordinateReferenceSystem crs = ...; // CRS obtained by any means.
crs = AbstractCRS.castOrCopy(crs).forConvention(AxesConvention.RIGHT_HANDED)

Coordinate Reference Systems

How do I instantiate a Universal Transverse Mercator (UTM) projection?

If the UTM zone is unknown, an easy way is to invoke the universal(…) method on one of the CommonCRS pre-defined constants. That method receives in argument a geographic coordinate in (latitude, longitude) order and computes the UTM zone from it. See the above Java code example.

If the UTM zone is know, one way is to use the “EPSG” or “AUTO” authority factory. The EPSG code of some UTM projections can be determined as below, where zone is a number from 1 to 60 inclusive (unless otherwise specified):

  • WGS 84 (northern hemisphere): 32600 + zone
  • WGS 84 (southern hemisphere): 32700 + zone
  • WGS 72 (northern hemisphere): 32200 + zone
  • WGS 72 (southern hemisphere): 32300 + zone
  • NAD 83 (northern hemisphere): 26900 + zone (zone 1 to 23 only)
  • NAD 27 (northern hemisphere): 26700 + zone (zone 1 to 22 only)

Note that the above list is incomplete. See the EPSG database for additional UTM definitions (WGS 72BE, SIRGAS 2000, SIRGAS 1995, SAD 69, ETRS 89, etc., most of them defined only for a few zones). Once the EPSG code of the UTM projection has been determined, the CRS can be obtained as in the example below:

int code = 32600 + zone;    // For WGS84 northern hemisphere
CoordinateReferenceSystem crs = CRS.forCode("EPSG:" + code);

How do I instantiate a Google projection?

The Google projection is a Mercator projection that pretends to be defined on the WGS84 datum, but actually ignores the ellipsoidal nature of that datum and uses the simpler spherical formulas instead. Since version 6.15 of EPSG geodetic dataset, the preferred way to get that projection is to invoke CRS.forCode("EPSG:3857"). Note that the use of that projection is not recommended, unless needed for compatibility with other data.

The EPSG:3857 definition uses a map projection method named “Popular Visualisation Pseudo Mercator”. The EPSG geodetic dataset provides also some other map projections that use spherical formulas. Those methods have “(Spherical)” in their name, for example “Mercator (Spherical)” (which differs from “Popular Visualisation Pseudo Mercator” by the use of a more appropriate sphere radius). Those projection methods can be used in Well Known Text (WKT) definitions.

If there is a need to use spherical formulas with a projection that does not have a “(Spherical)” counterpart, this can be done with explicit declarations of "semi_major" and "semi_minor" parameter values in the WKT definition. Those parameter values are usually inferred from the datum, but Apache SIS allows explicit declarations to override the inferred values.

How can I identify the projection kind of a CRS?

The “kind of projection” (Mercator, Lambert Conformal, etc.) is called Operation Method in ISO 19111 terminology. One approach is to check the value of OperationMethod.getName() and compare them against the OGC or EPSG names listed in the Coordinate Operation Methods page.

How do I get the EPSG code of an existing CRS?

The identifier of a Coordinate Reference System (CRS) object can be obtained by the getIdentifiers() method, which usually return a collection of zero or one element. If the CRS has been created from a Well Known Text (WKT) parsing and the WKT ends with an AUTHORITY["EPSG", "xxxx"] (WKT version 1) or ID["EPSG", xxxx] (WKT version 2) element, then the identifier (an EPSG numerical code in this example) is the xxxx value in that element. If the CRS has been created from the EPSG geodetic dataset (for example by a call to CRS.forCode("EPSG:xxxx")), then the identifier is the xxxx code given to that method. If the CRS has been created in another way, then the collection returned by the getIdentifiers() method may or may not be empty depending if the program that created the CRS took the responsibility of providing identifiers.

If the collection of identifiers is empty, the most effective fix is to make sure that the WKT contains an AUTHORITY or ID element (assuming that the CRS was parsed from a WKT). If this is not possible, then the org.apache.sis.referencing.IdentifiedObjects class contains some convenience methods which may help. In the following example, the call to lookupEPSG(…) will scan the EPSG database for a CRS equals (ignoring metadata) to the given one. Note that this scan is sensitive to axis order. Most geographic CRS in the EPSG database are declared with (latitude, longitude) axis order. Consequently if the given CRS has (longitude, latitude) axis order, then the scan is likely to find no match.

CoordinateReferenceSystem myCRS = ...;
Integer identifier = IdentifiedObjects.lookupEPSG(myCRS);
if (identifier != null) {
    System.out.println("The EPSG code has been found: " + identifier);

How do I get the “urn:ogc:def:crs:…” URN of an existing CRS?

OGC defines URN for CRS identifiers, for example "urn:​ogc:​def:​crs:​epsg:​7.1:​4326" where "7.1" is the version of the EPSG database used. URN may or may not be present in the set of identifiers returned by crs.getIdentifiers(). In many cases (especially if the CRS was parsed from a Well Known Text), only simple identifiers like "EPSG:​4326" are provided. An easy way to build the full URN is to use the code below. That example may scan the EPSG database for finding the information if it was not explicitely provided in the given CRS.

CoordinateReferenceSystem myCRS = ...;
String urn = IdentifiedObjects.lookupURN(myCRS);

Is IdentifiedObjects.lookupEPSG(…) a reliable inverse of CRS.forCode(…)?

For CRS created from the EPSG geodetic dataset, usually yes. Note however that IdentifiedObjects.getIdentifier(…) is cheaper and insensitive to the details of CRS definition, since it never query the database. But it works only if the CRS declares explicitly its code, which is the case for CRS created from the EPSG database or parsed from a Well Known Text (WKT) having an AUTHORITY or ID element. The lookupEPSG(…) method on the other hand is robust to erroneous code declaration, since it always compares the CRS with the database content. But it may fail if there is slight mismatch (for example rounding errors in projection parameters) between the supplied CRS and the CRS found in the database.

How can I determine if two CRS are “functionally” equal?

Two Coordinate Reference Systems may not be considered equal if they are associated to different metadata (name, identifiers, scope, domain of validity, remarks), even though they represent the same logical CRS. In order to test if two CRS are functionally equivalent, use Utilities​.equals­Ignore­Metadata(myFirstCRS, mySecondCRS).

Are CRS objects safe for use as keys in HashMap?

Yes, every classes defined in the, cs and datum packages define properly their equals(Object) and hashCode() methods. The Apache SIS library itself uses CRS objects in HashMap-like containers for caching purpose.

Coordinate transformations

My transformed coordinates are totally wrong!

This is most frequently caused by coordinate values given in the wrong order. Developers tend to assume a (x, y) or (longitude, latitude) axis order. But geographers and pilots are using (latitude, longitude) axis order for centuries, and the EPSG geodetic dataset defines geographic Coordinate Reference Systems that way. If a coordinate transformation seems to produce totally wrong values, the first thing to do should be to print the source and target Coordinate Reference Systems:


Attention should be paid to the order of AXIS elements. In the example below, the Coordinate Reference System clearly uses (latitude, longitude) axis order:

GeodeticCRS["WGS 84",
  Datum["World Geodetic System 1984",
    Ellipsoid["WGS 84", 6378137.0, 298.257223563]],
  CS[ellipsoidal, 2],
    Axis["Geodetic latitude (Lat)", north],
    Axis["Geodetic longitude (Lon)", east],
    Unit["degree", 0.017453292519943295]]

If (longitude, latitude) axis order is really wanted, Apache SIS can be forced to that order as described above.

I have correct axis order but my transformed coordinates are still wrong.

Make sure that the right projection is used. Some projection names are confusing. For example “Oblique Mercator” and “Hotine Oblique Mercator” (in EPSG naming) are two different projections. But “Oblique Mercator” (not Hotine) in EPSG naming is also called “Hotine Oblique Mercator Azimuth Center” by ESRI, while “Hotine Oblique Mercator” (EPSG naming) is called “Hotine Oblique Mercator Azimuth Natural Origin” by ESRI.

The “Oblique Stereographic” projection (EPSG name) is called “Double Stereographic” by ESRI. ESRI also defines a “Stereographic” projection, which is actually an oblique projection like the former but using different formulas.

I just used the WKT of a well-known authority and my transformed coordinates are still wrong!

The version 1 of Well Known Text (WKT) specification has been interpreted in different ways by different implementors. One subtle issue is the angular units of prime meridian and projection parameter values. The WKT 1 specification clary states: “If the PRIMEM clause occurs inside a GEOGCS, then the longitude units will match those of the geographic coordinate system” (source: OGC 01-009). However ESRI and GDAL among others unconditionally use decimal degrees, ignoring this part of the WKT 1 specification (note: this remark does not apply to WKT 2). This problem can be identified by WKT inspection as in the following extract:

PROJCS["Lambert II étendu",
  GEOGCS["Nouvelle Triangulation Française", ...,
    PRIMEM["Paris", 2.337229167],
    UNIT["grad", 0.01570796326794897]]
  PARAMETER["latitude_of_origin", 46.8], ...]

The Paris prime meridian is located at approximatively 2.597 gradians from Greenwich, which is 2.337 degrees. From this fact, we can see that the above WKT uses decimal degrees despite its UNIT["grad"] declaration. This mismatch applies also to the parameter value, which declare 46.8° in the above example while the official value is 52 gradians. By default, Apache SIS interprets those angular values as gradians when parsing such WKT, resulting in a large error. In order to get the intended result, there is a choice:

  • Replace UNIT["grad", 0.01570796326794897] by UNIT["degree", 0.017453292519943295], which ensure that Apache SIS, GDAL and ESRI understand that WKT 1 in the same way.

  • Or ask explicitely Apache SIS to parse the WKT using the ESRI or GDAL conventions, by specifying the Convention.WKT1_COMMON_UNITS enumeration value to WKTFormat in the package.

Note that the GeoPackage standard explicitely requires OGC 01-009 compliant WKT and the new WKT 2 standard also follows the OGC 01-009 interpretation. The default Apache SIS behavior is consistent with those two standards.

I verified all the above and still have an error of about one kilometer.

Coordinate Reference Systems (CRS) approximate the Earth’s shape by an ellipsoid. Different ellipsoids (actually different datum) are used in different countries of the world and at different time in history. When transforming coordinates between two CRS using the same datum, no Bursa-Wolf parameters are needed. But when the transformation involves a change of datum, the referencing module needs some information about how to perform that datum shift.

There is many way to specify how to perform a datum shift, and most of them are only approximation. The Bursa-Wolf method is one of them, not the only one. However it is one of the most frequently used methods. The Bursa-Wolf parameters can be specified inside a TOWGS84 element with version 1 of Well Known Text (WKT) format, or in a BOUNDCRS element with version 2 of WKT format. If the CRS are parsed from a WKT string, make sure that the string contains the appropriate element.

I get slightly different results depending on the environment I’m running in.

The results of coordinate transformations when running in a web application container (JBoss, etc.) may be a few meters off compared to coordinates transformations in an IDE (NetBeans, Eclipse, etc.). The results depend on whether an EPSG factory is available on the classpath, regardless how the CRS were created, because the EPSG factory specifies explicitly the coordinate operation to apply for some pairs of CRS. In such case, the coordinate operation specified by EPSG has precedence over the Burwa-Wolf parameters (the TOWGS84 element in version 1 of Well Known Text format).

A connection to the EPSG database may have been established for one environment (typically the JEE one) and not the other (typically the IDE one) because only the former has JDBC driver. The recommended way to uniformize the results is to add in the second environment (IDE) the same JDBC driver than the one available in the first environment (JEE). It should be one of the following: JavaDB (a.k.a. Derby), HSQL or PostgreSQL. Make sure that the connection parameters to the EPSG database are also the same.

Can I always expect a transform from an arbitrary CRS to WGS84 to succeed?

For 2D horizontal CRS created from the EPSG database, calls to CRS.findOperation(…) should generally succeed. For 3D CRS having any kind of height different than ellipsoidal height, or for a 2D CRS of type EngineeringCRS, it may fail. Note however that even if the call to CRS.findOperation(…) succeed, the call to MathTransform.transform(…) may fail or produce NaN or infinity values if the coordinate to transform is far from the domain of validity.


Custom implementations

My metadata are stored in a database-like framework. Implementing every GeoAPI interfaces for them is impractical.

Developers do not need to implement directly the metadata interfaces. If the underlying storage framework can access metadata from their class and attribute names (either Java names or ISO/OGC names), then it is possible to implement a single engine accessing any kind of metadata and let the Java Virtual Machine implements the GeoAPI interfaces on-the-fly, using the java.lang.reflect.Proxy class. See the Proxy Javadoc for details, keeping in mind that the ISO/OGC name of a java.lang.Class or java.lang.reflect.Method object can be obtained as below:

UML uml = method.getAnnotation(UML.class);
if (uml != null) {
    String name = uml.identifier();
    // Fetch the metadata here.

This is indeed the approach taken by the org.apache.sis.metadata.sql package for providing an implementation of all GeoAPI metadata interfaces reading their values directly from a SQL database.

I can not marshall my custom implementation.

The classes given to the JAXB marshaller shall contain JAXB annotations, otherwise the following exception is thrown:

javax.xml.bind.JAXBException: class MyCustomClass nor any of its super class is known to this context.

The easiest workaround is to wrap the custom implementation into one of the implementations provided in the org.apache.metadata.iso package. All those SIS implementation classes provide shallow copy constructor for making that easy. Note that you need to wrap only the root class, not the attributes. The attribute values will be wrapped automatically as needed by JAXB adapters.