Warn about (and fix) invalid DOIs

NEWS.md

@@ -9,6 +9,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ### Fixed
 
 * Warn about markdown link syntax in `.bib` files [[#60][]]
+* Warn about invalid DOIs in `.bib` files. The DOI field should never contain a URL (`https://doi.org/...`). This is detected as a special case, and the DOI is extracted from the URL.
 
 
 ## [Version 1.3.1][1.3.1] - 2023-11-02

docs/src/refs.bib

@@ -427,7 +427,7 @@ @incollection{TannorBookChapter1991
     pages = {333--345},
     publisher = {Springer},
     title = {Design of Femtosecond Pulse Sequences to Control Photochemical Products},
-    doi = {0.1007/978-94-011-2642-7_23},
+    doi = {10.1007/978-94-011-2642-7_23},
     year = {1991},
 }
 

src/formatting.jl

@@ -203,11 +203,12 @@ function format_published_in(
             push!(urls, doi_url(entry))
         end
     else
-        append!(urls, get_urls(entry))
-        if !isempty(get_title(entry))
+        if isempty(get_title(entry))
+            append!(urls, get_urls(entry; skip=0))
+        else
             # The title already linked to the URL (or DOI, if no
-            # URL available)
-            pop_url!(urls)
+            # URL available), hence we skip the first link here.
+            append!(urls, get_urls(entry; skip=1))
         end
     end
     segments = [String[], String[], String[]]
@@ -347,15 +348,22 @@ function get_booktitle(entry)
 end
 
 
-function get_urls(entry)
+function get_urls(entry; skip=0)
     # URL is first priority, DOI second (cf. `pop_url!`)
+    # Passing `skip = 1` skips the first available link
     urls = String[]
-    if !isempty(entry.access.doi)
-        url = doi_url(entry)
-        push!(urls, url)
-    end
     if !isempty(entry.access.url)
-        push!(urls, entry.access.url)
+        if skip <= 0
+            push!(urls, entry.access.url)
+        end
+        skip = skip - 1
+    end
+    if !isempty(entry.access.doi)
+        if skip <= 0
+            url = doi_url(entry)
+            push!(urls, url)
+        end
+        skip = skip - 1
     end
     return urls
 end
@@ -366,16 +374,28 @@ function doi_url(entry)
     if isempty(doi)
         return ""
     else
+        if !startswith(doi, "10.")
+            doi_match = match(r"\b10.\d{4,9}/.*\b", doi)
+            if isnothing(doi_match)
+                @warn "Invalid DOI $(repr(doi)) in bibtex entry $(repr(entry.id)). Ignoring DOI."
+                return ""
+            else
+                if startswith(doi, "http")
+                    @warn "The DOI field in bibtex entry $(repr(entry.id)) should not be a URL. Extracting $(repr(doi)) -> $(repr(doi_match.match))."
+                else
+                    @warn "Invalid DOI $(repr(doi)) in bibtex entry $(repr(entry.id)). Extracting $(repr(doi_match.match))."
+                end
+                doi = doi_match.match
+            end
+        end
         return "https://doi.org/$doi"
     end
 end
 
 
 function pop_url!(urls)
-    # pop from the back so later entries have higher priority (i.e., URL has
-    # highest priority if `urls=get_urls(entry)`.)
     try
-        return pop!(urls)
+        return popfirst!(urls)
     catch
         return ""
     end

test/test_formatting.jl

@@ -5,6 +5,8 @@ using DocumenterCitations
 import DocumenterCitations:
     two_digit_year,
     alpha_label,
+    get_urls,
+    doi_url,
     format_names,
     format_citation,
     format_bibliography_reference,
@@ -155,6 +157,20 @@ end
 end
 
 
+@testset "get_urls" begin
+    bib = CitationBibliography(DocumenterCitations.example_bibfile)
+    @test get_urls(bib.entries["GoerzSPIEO2021"]) == [
+        "https://michaelgoerz.net/research/GoerzSPIEO2021.pdf",
+        "https://doi.org/10.1117/12.2587002"
+    ]
+    @test get_urls(bib.entries["GoerzSPIEO2021"]; skip=1) ==
+          ["https://doi.org/10.1117/12.2587002"]
+    @test length(get_urls(bib.entries["GoerzSPIEO2021"]; skip=2)) == 0
+    @test get_urls(bib.entries["Nolting1997Coulomb"]) ==
+          ["https://doi.org/10.1007/978-3-663-14691-9"]
+end
+
+
 @testset "format_bibliography_reference(:numeric)" begin
     bib = CitationBibliography(DocumenterCitations.example_bibfile)
     md(key) = format_bibliography_reference(Val(:numeric), bib.entries[key])
@@ -367,3 +383,37 @@ end
 
 
 end
+
+
+@testset "invalid DOI" begin
+
+    bib = CitationBibliography(joinpath(splitext(@__FILE__)[1], "invalid_doi.bib"))
+
+    c = IOCapture.capture() do
+        doi_url(bib.entries["Brif"])
+    end
+    @test contains(
+        c.output,
+        "Warning: The DOI field in bibtex entry \"Brif\" should not be a URL."
+    )
+    @test c.value == "https://doi.org/10.1088/1367-2630/12/7/075008"
+
+    c = IOCapture.capture() do
+        doi_url(bib.entries["Shapiro"])
+    end
+    @test contains(
+        c.output,
+        "Warning: Invalid DOI \"doi:10.1002/9783527639700\" in bibtex entry \"Shapiro\"."
+    )
+    @test c.value == "https://doi.org/10.1002/9783527639700"
+
+    c = IOCapture.capture() do
+        doi_url(bib.entries["Tannor"])
+    end
+    @test contains(
+        c.output,
+        "Warning: Invalid DOI \"0.1007/978-94-011-2642-7_23\" in bibtex entry \"Tannor\"."
+    )
+    @test c.value == ""
+
+end

test/test_formatting/invalid_doi.bib

@@ -0,0 +1,38 @@
+@string{njp = "New J. Phys."}
+@string{XXXclearparser = ""}
+
+
+% DOI shouldn't be a URL
+@article{Brif,
+    Author = {Brif, Constantin and Chakrabarti, Raj and Rabitz, Herschel},
+    Title = {Control of quantum phenomena: past, present and future},
+    Journal = njp,
+    Year = {2010},
+    Doi = {https://doi.org/10.1088/1367-2630/12/7/075008},
+    Pages = {075008},
+    Volume = {12},
+}
+
+
+% DOI has extra test
+@book{Shapiro,
+    author = {Shapiro, Moshe and Brumer, Paul},
+    edition = {Second},
+    publisher = {Wiley and Sons},
+    title = {Quantum Control of Molecular Processes},
+    year = {2012},
+    Doi = {doi:10.1002/9783527639700},
+}
+
+
+% DOI is missing first letter
+@incollection{Tannor,
+    author = {Tannor, David J. and Jin, Yijian},
+    booktitle = {Mode Selective Chemistry},
+    editor = {Jortner, J. and Levine, R. D. and Pullman, B.},
+    pages = {333--345},
+    publisher = {Springer},
+    title = {Design of Femtosecond Pulse Sequences to Control Photochemical Products},
+    doi = {0.1007/978-94-011-2642-7_23},
+    year = {1991},
+}