fix(bib): a PubMed entry was reported as having no identifier; it has a PMID (#500)

CHANGELOG.md

@@ -366,6 +366,51 @@ flag changes and `doiget-mcp` tool spec changes will be called out explicitly he
   Also: `disposition` was inserted at six envelope sites and asserted at two -
   deleting one insert failed no test. The batch site now asserts it.
 
+- **[bib]** A PubMed-exported bibliography entry was reported as having **no
+  identifier**. It has a PMID (#500).
+
+  `entry has no DOI / arXiv id` is accurate about what the parser did and wrong
+  about the entry. A user reading it goes and edits a `.bib` that was fine; the
+  missing piece is on doiget's side. An existing test pinned exactly that claim,
+  with the comment "the entry has no resolvable identifier" - about an entry
+  carrying `eprinttype = {pubmed}`.
+
+  Both PubMed shapes are now named: the `pmid = {...}` field that PubMed's own
+  BibTeX export writes, and the BibLaTeX `eprint` + `eprinttype = {pubmed}` pair
+  that `arxiv_eligible` already refused, correctly and until now silently.
+  `pmcid` too.
+
+  It surfaces as **`NOT_IMPLEMENTED`, not `INVALID_REF`** - the input is valid
+  and the support is absent, and the two carry opposite advice ("wait for a
+  release" versus "correct your input"). Its disposition is `terminal`
+  accordingly (ADR-0055).
+
+  An entry with genuinely no identifier still reports `NoIdentifier`, and a DOI
+  alongside a PMID still wins: the check runs only after every supported
+  identifier has been tried, so it cannot divert an entry doiget could have
+  resolved.
+
+  This is the reporting half of #500. `Ref::Pmid` itself is blocked on something
+  the issue did not anticipate - see below.
+
+  The sentence the user reads lives in one place now
+  (`refs::unsupported_identifier_claim`). It had been hand-copied to the CLI
+  `verify` row and the MCP `batch_from_bibliography` envelope, and both copies
+  had been wrapped across source lines and re-joined with the indentation still
+  in them - shipping `which doiget                cannot resolve yet` to a
+  reader. Nothing asserted the text, so nothing failed. `#[error]` already
+  carried the same claim; the copies existed only to drop the `entry_key`
+  prefix, which each site puts in a field of its own.
+
+  Three more of the same, pre-existing and found while looking: two `= note:`
+  lines in `fetch fetch`'s widening advice and the `config.toml could not be
+  read` warning in `commands/mod.rs`. A workspace-wide lint for the pattern was
+  written and abandoned - it flags `config doctor`'s aligned two-column output
+  and test assertion messages at every space threshold from four to ten, and a
+  lint that cries wolf is a lint that gets deleted. Removing the duplication is
+  the durable half.
+
+
 - **[mcp]** `error.disposition` was **missing from five failure envelopes**,
   including the two most common failures an agent sees. The change that
   introduced it converted the envelopes built by `error_object` and missed the

Cargo.toml

@@ -15,7 +15,7 @@ exclude = [
 ]
 
 [workspace.package]
-version       = "0.8.13-beta.21"
+version       = "0.8.13-beta.22"
 edition       = "2021"
 rust-version  = "1.86"
 license       = "MIT"
@@ -41,9 +41,9 @@ features-doc-link = "docs/SOURCES.md"
 
 [workspace.dependencies]
 # Core
-doiget-core = { path = "crates/doiget-core", version = "0.8.13-beta.21" }
-doiget-cli  = { path = "crates/doiget-cli", version = "0.8.13-beta.21" }
-doiget-mcp  = { path = "crates/doiget-mcp", version = "0.8.13-beta.21" }
+doiget-core = { path = "crates/doiget-core", version = "0.8.13-beta.22" }
+doiget-cli  = { path = "crates/doiget-cli", version = "0.8.13-beta.22" }
+doiget-mcp  = { path = "crates/doiget-mcp", version = "0.8.13-beta.22" }
 
 # Async runtime — features are deliberately limited (avoid `full`).
 tokio = { version = "1", default-features = false, features = [

crates/doiget-cli/src/commands/batch.rs

@@ -121,6 +121,20 @@ pub async fn run_with_options(
         match entry {
             Ok(p) => inputs.push(p.ref_.as_input_str().to_string()),
             Err(ParseError::InvalidRef { raw, .. }) => inputs.push(raw),
+            // #500: a distinct placeholder, so the JSONL line says the entry
+            // HAS an identifier doiget cannot use rather than implying it has
+            // none. Same mechanism as the arm below; different claim.
+            Err(ParseError::UnsupportedIdentifier {
+                kind,
+                value,
+                entry_key,
+            }) => {
+                let placeholder = match entry_key {
+                    Some(k) => format!("<unsupported-{kind}:{value}:{k}>"),
+                    None => format!("<unsupported-{kind}:{value}>"),
+                };
+                inputs.push(placeholder);
+            }
             Err(ParseError::NoIdentifier { entry_key }) => {
                 // Synthesise a recognisable placeholder so Step 7's
                 // `Ref::parse` rejects this entry as `INVALID_REF`

crates/doiget-cli/src/commands/fetch.rs

@@ -1388,7 +1388,7 @@ fn not_found_trace_lines(ref_: &Ref, attempts: &[SourceAttempt]) -> Vec<String>
 
     if !already_set.is_empty() {
         out.push(format!(
-            "  = note: {} already set, but the source is still off -- this              binary was built without the Cargo feature that provides it.              Widening needs a differently-built binary, not another variable.",
+            "  = note: {} already set, but the source is still off -- this binary was built without the Cargo feature that provides it. Widening needs a differently-built binary, not another variable.",
             already_set.join(", ")
         ));
     }
@@ -1419,7 +1419,7 @@ fn not_found_trace_lines(ref_: &Ref, attempts: &[SourceAttempt]) -> Vec<String>
         // reading as an ordering it is not.
         if !middle.is_empty() {
             out.push(
-                "  = note: the middle is unordered because nothing in this run distinguishes                  those sources -- venue, affiliation and funder would, and none of them reach                  here. An invented order would read as information."
+                "  = note: the middle is unordered because nothing in this run distinguishes those sources -- venue, affiliation and funder would, and none of them reach here. An invented order would read as information."
                     .to_string(),
             );
         }

crates/doiget-cli/src/commands/mod.rs

@@ -231,7 +231,7 @@ fn store_root_from_config() -> Option<Utf8PathBuf> {
             tracing::warn!(
                 path = %path,
                 error = %e,
-                "config.toml could not be read; [store] root ignored and the default store                  root used instead. Run `doiget config doctor` to see which root is in effect."
+                "config.toml could not be read; [store] root ignored and the default store root used instead. Run `doiget config doctor` to see which root is in effect."
             );
             return None;
         }

crates/doiget-cli/src/commands/verify.rs

@@ -267,6 +267,27 @@ pub async fn run(path: String, format: String, cli_strict: bool, mode: OutputMod
                     "error": { "code": "INVALID_REF", "message": source.to_string() },
                 }),
             ),
+            // #500: still unverifiable, but for a reason the reader can act on
+            // -- and the action is not "fix the bibliography".
+            Err(ParseError::UnsupportedIdentifier {
+                kind,
+                value,
+                entry_key,
+            }) => (
+                VerifyStatus::Unverifiable,
+                serde_json::json!({
+                    "ok": false,
+                    "ref": serde_json::Value::Null,
+                    "status": VerifyStatus::Unverifiable.as_wire(),
+                    "entry_key": entry_key,
+                    "error": {
+                        "code": "NOT_IMPLEMENTED",
+                        "message": doiget_core::refs::unsupported_identifier_claim(
+                            kind, &value,
+                        ),
+                    },
+                }),
+            ),
             Err(ParseError::NoIdentifier { entry_key }) => (
                 VerifyStatus::Unverifiable,
                 serde_json::json!({