HDF5 reader: read variable-length string attributes and datasets

ReadAttrStr and HDF5DataSet::ReadString read every string as fixed-length,
sizing the buffer by H5Tget_size. For a variable-length string (what h5py
writes for a str, UTF-8) that size is a pointer, so the read put a char*
into the buffer and returned garbage. transformation_type then never equalled
"rotation", and an EIGER master whose goniometer attributes were written that
way (depends_on, transformation_type, units on /entry/sample/transformations)
was processed as stills although it holds a valid rotation sweep.

All string reads now go through one helper, ReadHDF5String, which handles both
fixed- and variable-length strings in either character set; ReadElement<string>
already handled the variable-length case and now uses the same helper.

Test: JFJochReader_Goniometer_VariableLengthStrings reads the rotation axis
from a master with fixed-length ASCII and with variable-length UTF-8 string
attributes and depends_on dataset.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SVmAWnzCmRKAXVUCdc4iNi

writer/HDF5Objects.cpp

@@ -506,6 +506,31 @@ HDF5Object & HDF5Object::Attr(const std::string &name, const std::vector<double>
     return *this;
 }
 
+std::string ReadHDF5String(hid_t type_id, const std::function<herr_t(hid_t, void *)> &read) {
+    if (H5Tis_variable_str(type_id) > 0) {
+        char *tmp = nullptr;
+        if (read(type_id, &tmp) < 0)
+            throw JFJochException(JFJochExceptionCategory::HDF5, "Cannot read variable-length string");
+        std::string out = (tmp != nullptr) ? std::string(tmp) : std::string{};
+        if (tmp != nullptr)
+            H5free_memory(tmp);
+        return out;
+    }
+
+    const size_t size = H5Tget_size(type_id);
+    if (size == 0)
+        return "";
+
+    std::string buffer(size, '\0');
+    if (read(type_id, buffer.data()) < 0)
+        throw JFJochException(JFJochExceptionCategory::HDF5, "Cannot read string");
+
+    const size_t end = buffer.find('\0');
+    if (end != std::string::npos)
+        buffer.resize(end);
+    return buffer;
+}
+
 std::vector<double> HDF5Object::ReadAttrVec(const std::string &name) {
     hid_t attr_id = H5Aopen(id, name.c_str(), H5P_DEFAULT);
     if (attr_id < 0)
@@ -579,11 +604,10 @@ std::string HDF5Object::ReadAttrStr(const std::string &name) {
         throw JFJochException(JFJochExceptionCategory::HDF5, "Attribute " + name + " is not a string");
     }
 
-    // Get string size and read it
-    size_t size = H5Tget_size(type_id);
-    std::vector<char> buffer(size + 1, '\0'); // +1 for null terminator to be safe
-
-    if (H5Aread(attr_id, type_id, buffer.data()) < 0) {
+    std::string ret;
+    try {
+        ret = ReadHDF5String(type_id, [&](hid_t type, void *buffer) { return H5Aread(attr_id, type, buffer); });
+    } catch (...) {
         H5Tclose(type_id);
         H5Aclose(attr_id);
         throw JFJochException(JFJochExceptionCategory::HDF5, "Cannot read attribute " + name);
@@ -591,7 +615,7 @@ std::string HDF5Object::ReadAttrStr(const std::string &name) {
 
     H5Tclose(type_id);
     H5Aclose(attr_id);
-    return {buffer.data()};
+    return ret;
 }
 
 
@@ -923,22 +947,13 @@ std::string HDF5DataSet::ReadString() const {
     }
 
     HDF5DataType file_data_type(*this);
-    const size_t size = file_data_type.GetElemSize();
-
-    if (size == 0)
-        return "";
-
-    std::string buffer(size, '\0');
-
-    if (H5Dread(id, file_data_type.GetID(), H5S_ALL, H5S_ALL, H5P_DEFAULT, buffer.data()) < 0)
-        throw JFJochException(JFJochExceptionCategory::HDF5,
-                              dataset_name + ": read unsuccessful");
-
-    const size_t end = buffer.find('\0');
-    if (end != std::string::npos)
-        buffer.resize(end);
-
-    return buffer;
+    try {
+        return ReadHDF5String(file_data_type.GetID(), [&](hid_t type, void *buffer) {
+            return H5Dread(id, type, H5S_ALL, H5S_ALL, H5P_DEFAULT, buffer);
+        });
+    } catch (const JFJochException &) {
+        throw JFJochException(JFJochExceptionCategory::HDF5, dataset_name + ": read unsuccessful");
+    }
 }
 
 void HDF5DataSet::Close() {