Time Complexity Analysis

Time Complexity Analysis

Overview

The Cetamura Batch Ingest Tool performs recursive discovery, metadata parsing, packaging, and validation for two workflows:

• Photo
• Patent

Overall runtime remains dominated by file discovery and asset processing.

Photo Workflow

Discovery

• recursive scan of candidate files: O(N)
• grouping and validation of discovered sets: O(N)

Where:

• N = number of files scanned under the selected root

Processing

Per matched asset:

• XML parsing: O(X)
• image decode and TIFF conversion: O(M)
• ZIP creation: O(M)

Where:

• X = XML size
• M = image size / pixel work

For P photo assets, the dominant runtime is:

• O(P * M)

Validation

• ZIP enumeration and structure checks: O(Z)
• reconciliation against CSV rows: O(Z + C)

Where:

• Z = ZIP count
• C = CSV row count

Patent Workflow

Discovery

• recursive scan for patent XML, PDF, and manifest files: O(N)

Processing

Per patent record:

• XML parsing and identifier checks: O(X)
• local PDF match lookup: O(L)
• fallback PDF index lookup: O(1) average after index build
• ZIP creation: O(F)

Where:

• X = XML size
• L = local PDF candidates in the batch directory
• F = file copy/compression work for one PDF, one XML, and one manifest

If fallback roots are used:

• fallback PDF index build: O(R)

Where:

• R = number of PDFs scanned across configured fallback roots

Total patent processing is typically:

• O(N + R + P)

for normal packaging workloads, with ZIP I/O dominating in practice.

Space Complexity

Typical memory use is bounded by:

• discovery metadata structures: O(N)
• one working asset at a time during packaging
• temporary output-side scratch space for photo workflow only

The design remains memory-efficient because packaging is sequential and source files are not duplicated in memory beyond the current item being processed.

Practical Bottlenecks

In real runs, the major cost is usually not the Python control flow. It is:

• disk I/O
• TIFF conversion for photo assets
• recursive fallback PDF indexing when patent fallback roots are large

Summary

• photo workflow is dominated by image processing and behaves roughly like O(P * M)
• patent workflow is dominated by recursive discovery plus ZIP I/O and behaves roughly like O(N + R + P)
• validation overhead is linear and small relative to packaging work

Last updated: April 6, 2026