Portuguese pretérito-perfeito 1st person plural (nós) forms truncate to stems for many regular -ar verbs

[sshpiz]

(I used an AI agent to reproduce and write below. I hope this is useful and not unuseful.)

Version tested: verbecc 2.0.2 from PyPI

I can reproduce this in the library itself (not just in a downstream UI) for multiple regular Portuguese -ar verbs.

I also reproduced the same behavior against the current main branch on 2026-04-22 (commit 73480711fb18800533000e7f56ff7fcb4bbc964a).

Minimal reproduction:

from verbecc import CompleteConjugator, LangCodeISO639_1 as Lang, MoodPt, TensePt

verbs = ["contar", "pensar", "chegar", "passar", "deixar", "ajudar", "cantar", "comprar"]
cc = CompleteConjugator(Lang.pt)

for verb in verbs:
    forms = cc.conjugate(verb).get_moods()[MoodPt.Indicativo][TensePt.PretéritoPerfeito]
    print(verb, forms[5].get_data()["c"][0])

Observed output:

• contar -> nós cont
• pensar -> nós pens
• chegar -> nós cheg
• passar -> nós pass
• deixar -> nós deix
• ajudar -> nós ajud
• cantar -> nós cant
• comprar -> nós compr

Expected output:

• contar -> nós contámos
• pensar -> nós pensámos
• chegar -> nós chegámos
• passar -> nós passámos
• deixar -> nós deixámos
• ajudar -> nós ajudámos
• cantar -> nós cantámos
• comprar -> nós comprámos

So far this looks broader than a single verb and at least affects regular -ar templates.

I inspected the Portuguese conjugation XML in the installed package and noticed that the 4th slot of pretérito-perfeito is empty in at least these templates:

• compr:ar
• cheg:ar

For example, the pretérito-perfeito block contains an empty 4th <p><i/></p> entry where the nós form should be derived.

That seems consistent with the truncated outputs above.

As extra local investigation context: in an older vendored copy I also noticed malformed Portuguese template values such as ) and r) in related areas, but the current 2.0.2 package already reproduces the bug with the empty pretérito-perfeito slot(s) above.