feat(core): flag "tenant" misused for "tenet"

[ozpool]

Issues

Closes #1070.

Description

tenant (a renter) is commonly written where tenet (a principle) is meant — "a core tenant of good design", "the main tenants of many religions". Harper did not catch this.

This adds a TenantTenet rule that flags tenant / tenants when it follows a principle-sense adjective (core, central, fundamental, basic, guiding, cardinal, underlying, foundational, defining) and suggests tenet / tenets:

• a core tenant of good design → a core tenet of good design
• the central tenants of the movement → the central tenets of the movement

To keep false positives low, it deliberately only fires after that adjective set, where the renter reading is implausible. Real-estate senses are left untouched — the main tenant, a key tenant, three tenants, commercial tenants. Broader frames (e.g. bare tenants of X) are intentionally out of scope for now, since tenant is a valid word and over-flagging would be worse than the occasional miss.

The rule is a grouped Weir rule (singular + plural children under TenantTenet/) because the two forms need different replacements.

No related PRs.

Demo

$ harper-cli lint "These are the core tenants of our design."
1 │ These are the core tenants of our design.
  │                   ╰──────── [Malapropism::TenantTenet]: Did you mean `tenets`?

$ harper-cli lint "The main tenant of the building left."   # no lint (renter)

How Has This Been Tested?

Embedded Weir tests (16 correction cases + 8 negative cases across the two children):

$ harper-cli test harper-core/src/linting/weir_rules/TenantTenet/Singular.weir
All tests pass!
$ harper-cli test harper-core/src/linting/weir_rules/TenantTenet/Plural.weir
All tests pass!

Rust suite (rule registration + curated-config drift check):

$ cargo test -p harper-core --lib curated_default_config_lists_every_registered_rule
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 5803 filtered out

$ cargo test -p harper-core --lib run_tests_for_weir_rules
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 5803 filtered out

Also checked by hand that renter-sense phrases (the main tenant, a key tenant, three tenants) stay clean.

AI Disclosure

• I am a human and didn't use any AI.
• I used LLM features of my editor, but not an agent.
• I used an AI agent interactively.
• I am an agent or I got an agent to do the work autonomously.

If Your PR Implements or Enhances a Linter

• I made up the sentences in the unit tests.
• The sentences in the unit tests were generated by an AI.
• I'm using examples from the bug report / feature request.
• I collected real-world sentences for the unit tests.

Checklist

• I have performed a self-review of my own code
• I have added tests to cover my changes
• I have considered splitting this into smaller pull requests.

[hippietrail]

(I wrote this message hours ago while on the road, but it seems as though I failed to submit it?)

I would like to see this run over a large corpus of real-world text somehow. Mainly because when I was researching this with tons of manual Google searches I found "tenant" used legitimately in tech contexts far more than I was expecting. So much so that it proved very difficult to find examples of the mistake.

By the way, I've never heard or read anybody using "tenet" when they intend the meaning of "tenant" so we don't need to handle that case.

Now I've tried quite a lot to get an AI agent to harvest real-world examples for Harper testing and so far I've not found anything that works. In particular I can't get them to automate my way, which is Google searches with search operators to include all inflections and sometimes wildcards. But I've only used free AI tools. Maybe somebody with access to paid tools who is interested in tackling this can do it.

[ozpool]

Good call — you were right, and it did fire on the tech usage you mentioned. Going looking for it, I hit exactly that:

• "The core tenant isolation logic lives here."
• "Our fundamental tenant model uses row-level security."
• "The basic tenant configuration is in config.yaml."

The problem is that in multi-tenancy writing tenant is a noun modifier, so the adjective binds to the whole compound — core [tenant isolation], not [core tenant]. The adjective gate on its own couldn't tell those apart.

Pushed a fix: the rule now requires the token after tenant / tenants to be a non-noun, which leaves the compound cases alone while still catching the mistake.

text	before	after
core tenant isolation logic	flagged	clean
fundamental tenant model	flagged	clean
basic tenant configuration	flagged	clean
core tenants list is stored per region	flagged	clean
multi-tenant architecture, tenant ID	clean	clean
the tenant paid rent, commercial tenants	clean	clean
a core tenant of good design	flagged	flagged
one of my central tenants is honesty	flagged	flagged
Core Tenants of Highly Effective Teams	flagged	flagged

One trade-off worth flagging: a sentence-final a core tenant. is no longer caught, since there is no following token to inspect. Avoiding the false positives seemed worth more than catching that form.

Agreed on the direction, too — the rule only goes tenant → tenet, never the reverse.

On the corpus point: I don't have a good automated harvesting setup either. What I have done instead is keep the trigger deliberately narrow — only the nine principle-sense adjectives, plus the noun-modifier exclusion — so it stays conservative and misses real instances (key tenants, main tenants) rather than risk firing on legitimate text. Happy to widen it if a corpus ever shows that is safe.