Skip to main content

Performance

CIMXML is built for large power-system models. It streams parsing rather than buffering whole documents and applies difference models as deltas instead of materializing copies. This page covers the choices that matter when you process big files.

Memory optimization​

Parsed graphs use plain Apache Jena GraphFactory.createGraphMem() graphs. The one specialized mechanism the module adds on top is FastDeltaGraph, which applies difference models as a delta over the base graph without materializing a merged copy.

Difference application without copies​

Applying a difference model with differenceModelToFullModel(...) returns a FastDeltaGraph layered over the predecessor body. Additions and removals are held in their own GraphFactory.createGraphMem() delta graphs rather than rewriting the base, so applying a difference to a large model stays cheap in both time and memory. See Difference models.

Large file handling​

When you parse from a Path, CIMXML reads through a BufferedFileChannelInputStream with a buffer sized to the file — capped at a maximum so very large files do not allocate an oversized buffer:

// Parsing from a Path uses a buffered file channel internally
Path largeCimFile = Path.of("large_model.xml");
CimDatasetGraph dataset = parser.parseCimModel(largeCimFile);

For smaller inputs the buffer matches the file size; beyond the internal maximum it is clamped to a fixed size. You do not configure this — it is chosen automatically by the Path overload.

Prefer the Path overload for files

Passing a Path lets the library pick an optimal buffered channel and size it for you. Use the InputStream / Reader overloads for in-memory or streamed sources where you already control buffering.

Reuse the parser

A single CimXmlParser is thread-safe for parsing and holds the profile registry, so register your profiles once and reuse the parser across many model files rather than recreating it per file.