Adds TStreamerSTLstring support (#378)

* Adds TStreamerSTLstring support

* Fix readout for STLstring

* Add tests

* Fix typo

Author: tamasgal

src/custom.jl

@@ -30,11 +30,11 @@ end
 
 
 # running length coded string
-function runlength_string(::Type{T}, data) where T
+function runlength_string(::Type{T}, data; offset=10) where T
     out = T[]
     maxlength = length(data)
-    # offsetjagg skip 10 bytes
-    i = 11
+    # offsetjagg skip
+    i = offset + 1
     while true
         i+1 > maxlength && break
         _len = data[i]
@@ -46,10 +46,10 @@ function runlength_string(::Type{T}, data) where T
     out
 end
 
-function interped_data(data::Vector{UInt8}, rawoffsets::Vector{Int32}, ::Type{Vector{T}}, ::Type{Offsetjagg}) where {T<:AbstractString}
+function interped_data(data::Vector{UInt8}, rawoffsets::Vector{Int32}, ::Type{Vector{T}}, ::Type{J}) where {T<:AbstractString, J<:Union{Offsetjagg, Offset6jagg}}
     rawoffsets .+= 1
     v = VectorOfVectors(data, rawoffsets)
-    res = runlength_string.(T, v)
+    res = runlength_string.(T, v; offset=offsetof(J))
     dummy = VectorOfVectors(res)
     # to maintain Int32 indexing
     return VectorOfVectors(dummy.data, Vector{Int32}(dummy.elem_ptr))

src/root.jl

@@ -418,6 +418,11 @@ function auto_T_JaggT(f::ROOTFile, branch; customstructs::Dict{String, Type})
                 streamer.fTypeName == "vector<unsigned char>" && return Vector{Vector{UInt8}}, _jaggtype
             end
 
+            # TODO: unclear if it's always a 6 byte offset but it solve issue #377
+            # also note that utils.jl (JaggType(f, branch, leaf)) might be the better
+            # place to determine jaggedness and offset size
+            streamer.fTypeName == "string" && return Vector{String}, Offset6jagg
+
         end
 
         # some standard cases

src/utils.jl

@@ -54,9 +54,12 @@ abstract type JaggType end
 struct Nojagg      <:JaggType  end
 struct Nooffsetjagg<:JaggType  end
 struct Offsetjagg  <:JaggType  end
+struct Offset6jagg  <:JaggType  end
 struct Offsetjaggjagg  <:JaggType  end
 # this is a preliminary workaround for 6 byte offset jaggedness
 struct Offset6jaggjagg  <:JaggType  end
+offsetof(::Type{Offsetjagg}) = 10
+offsetof(::Type{Offset6jagg}) = 6
 
 function JaggType(f, branch, leaf)
     # https://github.com/scikit-hep/uproot3/blob/54f5151fb7c686c3a161fbe44b9f299e482f346b/uproot3/interp/auto.py#L144
@@ -75,6 +78,11 @@ function JaggType(f, branch, leaf)
             # TODO: there are for sure also jagged strings, need to find files with those
             return Nojagg
         end
+        # if typeof(streamer) <: TStreamerSTLstring
+        #     # TODO: unclear if this jagged/offset info should be passed from here
+        #     # or dealt with in root.jl. see issue #377 and PR #378
+        #     return Offsetjagg
+        # end
         if streamer.fSTLtype == Const.kSTLvector
             (match(r"\[.*\]", leaf.fTitle) !== nothing) && return Offset6jaggjagg
             return Offsetjagg

test/issues.jl

@@ -57,6 +57,16 @@ SAMPLES_DIR = joinpath(@__DIR__, "samples")
     @test 1200 == length(t)
     @test t[1].time[2] ≈ 36.396744f0
     @test t[end].xpos[end] ≈ 788.35144f0
+
+    # issue 377
+    f = UnROOT.samplefile("issue377.root")
+    arr = UnROOT.array(f, "podio_metadata/events___CollectionTypeInfo/events___CollectionTypeInfo.dataType")
+    t = LazyTree(f, "podio_metadata", ["events___CollectionTypeInfo"])
+    @test 1 == length(t.events___CollectionTypeInfo_dataType)
+    @test 26 == length(t.events___CollectionTypeInfo_dataType[1])
+    @test "edm4hep::CaloHitContributionCollection" == t.events___CollectionTypeInfo_dataType[1][1]
+    @test "podio::LinkCollection<edm4hep::Vertex,edm4hep::ReconstructedParticle>" == t.events___CollectionTypeInfo_dataType[1][end]
+    @test arr == t.events___CollectionTypeInfo_dataType
 end
 
 function _test_clean_GC(fname)