Undocumented query syntax limits and parsing failures in govinfo.gov search

[Dorado1987A]

Hi team,

While testing advanced Boolean search behaviour, I encountered a series of consistent parser failures that don’t appear to be documented. I’d like to confirm whether these are known or intended constraints, or if they represent bugs in the current implementation.

---

Key Findings

1. Proximity + OR-lists with multi-word phrases

Queries using NEAR/# fail when either side includes an OR-list with multi-word items, even when written as ADJ chains.

Examples:

(("Direct Action Everywhere" OR PETA) NEAR/20 milk)          → returns 0 results
(((Direct ADJ Action ADJ Everywhere) OR PETA) NEAR/20 milk)  → also fails
(PETA) NEAR/20 (milk OR dairy)                               → works

2. Quoted tokens break proximity

Any quoted token on either side of NEAR/# breaks the query—even if it's a single word.

"cow" NEAR/20 abuse   → fails
cow NEAR/20 abuse     → works

3. OR-list size limits (based on quoting)

• Quoted tokens: fails when the list exceeds 30 items
• Unquoted tokens: works up to around 70 items

("term1" OR "term2" ... "term31")   → fails
(term1 OR term2 ... term50)         → works

4. Query complexity limit appears to be token-based

There is a hard cap on the number of single-word tokens a query can include, regardless of character count. Using a repeated structure like:

((cow OR calf) NEAR/20 (welfare OR abuse)) OR ...

• Works with 69 tokens (≈ 923 characters)
• Fails at 70 tokens
• Using ADJ (e.g. cow ADJ calf) lowers the cap to 66 tokens

This suggests a token or operator count limit rather than a character limit.

5. Wildcards inside quotes fail

"cow*"                 → fails
"animal welf*"         → fails
cow*                   → works

---

Request

Can you confirm whether these are known limitations of the current govinfo.gov search parser? If they are intended, clarification around the following would be appreciated:

• Allowed structure and token count for proximity clauses
• Safe limits for OR-lists (quoted vs unquoted)
• Whether multi-token phrases can be safely used near NEAR/# or ADJ
• Whether wildcard handling inside quotes is expected to fail

Happy to share a full reproducible test suite if that’s useful.

Thanks!

[Dorado1987A]

The error code I get is: {"message":"Oops, Something went wrong, Please contact govinfo team for further assistance"}

With no other details

[jonquandt]

Thanks for passing these along - We'll review these use cases and get back to you. To confirm, you're seeing this behavior in both the GovInfo API and the regular govinfo.gov search?

edit: if you want to pass along additional examples or test documentation you've developed, that may help with analysis on our end.

[Dorado1987A]

Hi @jonquandt, sorry I wasn't notified of you responding! Yes, I'm seeing this in both the API and the regular search on the website.

As for documentation, is there an email I can use or a way to send you it directly rather than posting here?

[jonquandt]

@Dorado1987A - I updated my profile to include my email.

[Dorado1987A]

@jonquandt emailed

[jonquandt]

Thanks for the additional information. We're reviewing and will provide a response to you individually and via GitHub with additional documentation updates. This will also likely result in updates to the GovInfo help pages surrounding search. Timing for this is TBD since we are in the midst of a particularly busy part of our current dev cycle.