[PATCH] MITAB: some French EPSG code doesn't seems to be supported

[yjacolin]

Here is an issue showed by a french user on GeoRezo Portail (​http://georezo.net/forum/viewtopic.php?pid=155606)

Hello,

When transforming a MapInfo File from one coordinate Reference System to the following EPSG codes EPSG:27584, 27574, 27572, the result is a reprojected file with a false header. The header contains "nonearth" information. When transforming into RGF93 (EPSG:2154), the result is a header that contains WGS84 srs projection information.

The issue is about the header writing for these coordinate reference systems. These are the only ones that I've tested until now.

Best,

Mathieu

[Even Rouault]

Daniel,

A quick analysis shows that all those EPSG codes are based on the "Lambert_Conformal_Conic_1SP" projection method, whereas the MITAB code only seems to handle LCC_2SP

For EPSG:2154, the expansion of the code gives "Reseau_Geodesique_Francais_1993" for the datum name. This datum is not listed in the asDatumInfoList[] array, so it defaults to using WGS84.

[yjacolin]

Sample data in MapInfo format

[yjacolin]

Hello,

I attach a zip file contening two mapinfo file, maTable.* is a sample dataset in Lambert 93 srs from Mapinfo and MATAble_EPSG2154 made by QGIS after opening the first dataset and save it to MapInfo file.

Hope it helps,

Y.

[Daniel Morissette]

The 'tabdump' program returns the following info from the header of the MaTable.MAP file in the attached ZIP:

  m_sProj.nDatumId      = 33
  m_sProj.nProjId       = 3
  m_sProj.nEllipsoidId  = 0
  m_sProj.nUnitsId      = 7
  m_sProj.adProjParams  = 3 46.5 44 49 700000 6.6e+06
  m_sProj.dDatumShiftX  = 0
  m_sProj.dDatumShiftY  = 0
  m_sProj.dDatumShiftZ  = 0
  m_sProj.adDatumParams = 0 0 0 0 0

ProjId == 3 is already handled as LCC on read by calling OGRSpatialReference::SetLCC() in mitab_spatialref.cpp's TABFile::GetSpatialRef(). How can we tell OGR to treat the projection as LCC_1SP instead of LCC_2SP which seems to happen by default?

[Even Rouault]

I've attached a patch to address the GRS_80 vs GRS_1980 issue, map Reseau_Geodesique_Francais_1993 to GRS_80 (since apparently there's no difference in the file between the 2 datums), and avoid 0 to be turned into -0 in TOWGS84 . This doesn't solve the LCC_1SP vs LCC_2SP issue.

[Even Rouault]

Patch to map Reseau_Geodesique_Francais_1993 to GRS_80, and avoid 0 to be turned into -0 in TOWGS84

[Even Rouault]

I've removed the GRS_1980 alias. I must have messed up something in my test. It is likely unnecessary.

[Even Rouault]

In 40854:

Add OGRSpatialReference::convertToOtherProjection() (bound to C as OSRConvertToOtherProjection() and to SWIG) to transform between different equivalent projections (currently Merc_1SP <--> Merc_2SP and LCC_1SP <--> LCC_2SP) (refs #3292)

[Even Rouault]

In 40855:

MITAB: accept creating files with LCC_1SP SRS (fixes #3292)