Repository navigation
NLP_ENGINE tries to add analyzer already in memory #380
Description
NLP_Engine tries to add existing analyzer, tries to read kb that is already in memory.
"Active Algo count=1" also means that an Algo data structure is not freed (ie memory leak).
stderr.log file, eg running CodedExamples, shows the below errors.
[addAna: Named analyzer already present: d:\git\mvr\analyzers\CodedExamples]
[con_add_root: Root concept of kb already exists.]
[bind_sys: Enter.]
[Loaded knowledge base.]
0 0 [Exec analyzer time=0.079 sec]
0 0 Active Algo count=1
Activity
fix: re-entrant analyze() rebuilds + orphans NLP on path-form mismatch (
ddehilster commentedon Jun 17, 2026
Fixed by #658 (engine 3.5.6). Same root cause as reported here:
NLP_ENGINE::initlooked up "is this analyzer already loaded?" by the raw caller path, but the NLP is registered under its basename, so on re-entry it failed to match — re-reading the KB and adding an analyzer already in memory, leaking the prior one (theActive Algo count=1symptom). The redundant explicitaddAna(the NLP constructor already registers) produced the[addAna: Named analyzer already present]log.#658 keys the registry on the canonical analyzer name so a re-entrant analyze() reuses the loaded NLP instead of rebuilding, and removes the redundant
addAna. Verified in VisualText/py-package-nlpengine#31: the re-entry now logs[Reusing loaded analyzer: ...]with no "already present" noise and no rebuild.