[
https://issues.apache.org/jira/browse/SIS-433?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Martin Desruisseaux updated SIS-433:
------------------------------------
Description:
In the EPSG geodetic database, units of measurement are defined by a _b_/_c_
ratio. But when parsing a WKT, the {{UNIT}} element is defined only by a scale
factor which is the result of above division. This result is a slight lost of
precision. Example with EPSG::9042 unit:
{quote}
British chain (Sears 1922) = (792 / 39.370147) metres ≈ 20.116765121552632
metres
{quote}
Which is formatted in WKT as below (note: we should opportunistically fix the
unit name, but it does not impact this issue):
{noformat}
Unit["20.116765121552632⋅m", 20.116765121552632]
{noformat}
Unit conversions in Apache SIS internally use ratios for better accuracy.
Consequently CRS objects created from the EPSG database get slightly different
unit conversion factors than equivalent CRS objects parsed from WKT. This very
small difference cascades in non-linear computations up to a point where two
CRS objects are considered not equal, even in {{ComparisonMode.APPROXIMATIVE}}
way (see SIS-434 for more in-depth description of that problem). The proposed
improvement is to use the EPSG code of coordinate parameter values for fetching
the unit of measurement from EPSG database. This work is described in SIS-210;
fixing that later issue should help fixing this issue.
h2. Test case for debugging
In
{{core/sis-referencing/src/test/java/org/apache/sis/test/integration/ConsistencyTest.java}},
put a {{\@Test}} annotation on the {{debug()}} method and temporarily disable
other tests.
If we want to opportunistically fix the unit name issue, it may be done in
{{EPSGDataAccess}} by invoking {{ConventionalUnit.forSymbol(String)}}, maybe
with the help of an interface defined in an internal package.
was:
In the EPSG geodetic database, units of measurement are defined by a _b_/_c_
ratio. But when parsing a WKT, the {{UNIT}} element is defined only by a scale
factor which is the result of above division. This result is a slight lost of
precision. Example with EPSG::9042 unit:
{quote}
British chain (Sears 1922) = (792 / 39.370147) metres ≈ 20.116765121552632
metres
{quote}
Unit conversions in Apache SIS internally use ratios for better accuracy.
Consequently CRS objects created from the EPSG database get slightly different
unit conversion factors than equivalent CRS objects parsed from WKT. This very
small difference cascades in non-linear computations up to a point where two
CRS objects are considered not equal, even in {{ComparisonMode.APPROXIMATIVE}}
way (see SIS-434 for more in-depth description of that problem). The proposed
improvement is to use the EPSG code of coordinate parameter values for fetching
the unit of measurement from EPSG database. This work is described in SIS-210;
fixing that later issue should help fixing this issue.
> Replace unit parsed from WKT by unit declared in EPSG database
> --------------------------------------------------------------
>
> Key: SIS-433
> URL: https://issues.apache.org/jira/browse/SIS-433
> Project: Spatial Information Systems
> Issue Type: Task
> Components: Referencing
> Reporter: Martin Desruisseaux
> Priority: Minor
>
> In the EPSG geodetic database, units of measurement are defined by a _b_/_c_
> ratio. But when parsing a WKT, the {{UNIT}} element is defined only by a
> scale factor which is the result of above division. This result is a slight
> lost of precision. Example with EPSG::9042 unit:
> {quote}
> British chain (Sears 1922) = (792 / 39.370147) metres ≈ 20.116765121552632
> metres
> {quote}
> Which is formatted in WKT as below (note: we should opportunistically fix the
> unit name, but it does not impact this issue):
> {noformat}
> Unit["20.116765121552632⋅m", 20.116765121552632]
> {noformat}
> Unit conversions in Apache SIS internally use ratios for better accuracy.
> Consequently CRS objects created from the EPSG database get slightly
> different unit conversion factors than equivalent CRS objects parsed from
> WKT. This very small difference cascades in non-linear computations up to a
> point where two CRS objects are considered not equal, even in
> {{ComparisonMode.APPROXIMATIVE}} way (see SIS-434 for more in-depth
> description of that problem). The proposed improvement is to use the EPSG
> code of coordinate parameter values for fetching the unit of measurement from
> EPSG database. This work is described in SIS-210; fixing that later issue
> should help fixing this issue.
> h2. Test case for debugging
> In
> {{core/sis-referencing/src/test/java/org/apache/sis/test/integration/ConsistencyTest.java}},
> put a {{\@Test}} annotation on the {{debug()}} method and temporarily
> disable other tests.
> If we want to opportunistically fix the unit name issue, it may be done in
> {{EPSGDataAccess}} by invoking {{ConventionalUnit.forSymbol(String)}}, maybe
> with the help of an interface defined in an internal package.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
