[FileFormats.MPS] unable to parse rvb-sub.mps

[ZedongPeng]

I was trying to read rvb-sub.mps using the read_from_file function. However, it seems that this instance uses a special MPS format with entries like E s0 $t5 0, which are not currently supported by JuMP.

Is there any plan to support this extended MPS syntax in future versions?

ERROR: LoadError: Malformed ROWS line: E s0 $t5 0
Stacktrace:
  [1] error(s::String)
    @ Base ./error.jl:35
  [2] parse_rows_line(data::MathOptInterface.FileFormats.MPS.TempMPSModel, items::Vector{String})
    @ MathOptInterface.FileFormats.MPS ~/.julia/packages/MathOptInterface/doGwJ/src/FileFormats/MPS/MPS.jl:1406
  [3] read!(io::IOStream, model::MathOptInterface.Utilities.GenericModel{Float64, MathOptInterface.Utilities.ObjectiveContainer{Float64}, MathOptInterface.Utilities.VariablesContainer{Float64}, MathOptInterface.FileFormats.MPS.ModelFunctionConstraints{Float64}})
    @ MathOptInterface.FileFormats.MPS ~/.julia/packages/MathOptInterface/doGwJ/src/FileFormats/MPS/MPS.jl:1207
  [4] jump-dev/JuMP.jl#11
    @ ~/.julia/packages/MathOptInterface/doGwJ/src/FileFormats/FileFormats.jl:150 [inlined]
  [5] open(::MathOptInterface.FileFormats.var"#11#12"{MathOptInterface.Utilities.GenericModel{Float64, MathOptInterface.Utilities.ObjectiveContainer{Float64}, MathOptInterface.Utilities.VariablesContainer{Float64}, MathOptInterface.FileFormats.MPS.ModelFunctionConstraints{Float64}}}, ::String, ::Vararg{String}; kwargs::@Kwargs{})
    @ Base ./io.jl:410
  [6] open
    @ ./io.jl:407 [inlined]
  [7] compressed_open
    @ ~/.julia/packages/MathOptInterface/doGwJ/src/FileFormats/utils.jl:282 [inlined]
  [8] compressed_open(f::MathOptInterface.FileFormats.var"#11#12"{MathOptInterface.Utilities.GenericModel{Float64, MathOptInterface.Utilities.ObjectiveContainer{Float64}, MathOptInterface.Utilities.VariablesContainer{Float64}, MathOptInterface.FileFormats.MPS.ModelFunctionConstraints{Float64}}}, filename::String, mode::String, ::MathOptInterface.FileFormats.AutomaticCompression)
    @ MathOptInterface.FileFormats ~/.julia/packages/MathOptInterface/doGwJ/src/FileFormats/utils.jl:320
  [9] read_from_file(model::MathOptInterface.Utilities.GenericModel{Float64, MathOptInterface.Utilities.ObjectiveContainer{Float64}, MathOptInterface.Utilities.VariablesContainer{Float64}, MathOptInterface.FileFormats.MPS.ModelFunctionConstraints{Float64}}, filename::String)
    @ MathOptInterface.FileFormats ~/.julia/packages/MathOptInterface/doGwJ/src/FileFormats/FileFormats.jl:149
 [10] read_from_file(filename::String; format::MathOptInterface.FileFormats.FileFormat, kwargs::@Kwargs{})
    @ JuMP ~/.julia/packages/JuMP/RGIK3/src/file_formats.jl:282
 [11] read_from_file
    @ ~/.julia/packages/JuMP/RGIK3/src/file_formats.jl:276 [inlined]

[odow]

E s0 $t5 0

What does this line mean? What is $t5 and what is the last 0?

[ZedongPeng]

I have no idea. Should I report this to miplib instead?

[odow]

@Opt-Mucca halp plz

[odow]

gurobi_cl appears to just ignore the extra fields... @simonbowly I missed this meeting also.

[simonbowly]

Not a clue, the Gurobi docs only show two entries here. I guess all the solvers are ignoring extra fields or the MIPLIB tests would have fallen over.

[Opt-Mucca]

I think the $t5 0 is just ignored by all solvers. Don't think it's even valid MPS format........ The $tX "rows" at least never feature in anything.

To clarify: The check we do for MIPLIB is "can all solvers read in the instance, and after reading is the instance identical". This instance apparently passes that while still being invalid.