Skip to content

elifesciences/enhanced-preprints-biorxiv-xslt

Repository files navigation

Documentation justifying each XSL file

Must contain an entry for each XSL file justifying the need for this. The more detail the better.

Documentation should include recommendations for how each XSL file could ultimately be removed by one of:

  • improved biorXiv XML
  • better handling of XML in Encoda
  • better support for Encoda output in EPP client

Please provide links to github issues which express these requirements.

In the entry for each XSL file please link to the file in github.

Global XSLT

This stylesheet is transforming an XML document by adding a "label" element to any "title" element that has a preceding "label" element, and removing any "label" element that has a following "title" element. Work is required in encoda to approprately decode these labels, so that they can then be represented in the resultant JSON.

The XSL also changes an label element (that is not followed by a title element) to a title element within sec and app elements.

This xsl converts <ext-link> elements within references (<ref>) that have a uri (in the xlink:href attribute) which is a DOI string, to use <pub-id pub-id-type="doi"> tagging instead.

<pub-id pub-id-type="doi"> is preferred capture, and this should be fed back to bioRxiv. In addition, encoda needs to be able to appropriately decode <ext-link> elements in references, so that these can be represented in the JSON, because it is otherwise perfectly acceptable capture in JATS (for example, when the link is not a DOI).

This xsl is a workaround for handling group authors (organisations) in both the author list and in references. Encoda converts these into a useful representation in the JSON, but there is no support for authors that are organisations in EPP client, so this needs updating.

This xsl handles multiple reference lists in a preprint. If the preprint has mutliple reference lists (an example of this is 10.1101/2022.12.20.521179), then the references from any extra reference lists are added to the first one. This means that the references are retained (although the separation of the reference lists and the extra headings cannot be retained).

Encoda is currently unable to handle to handke files with mutliple reference lists. This will need resolving first, becfore EPP client being updated, so that mutliple reference lists can be rendered with their headings.

Changes are required to encoda so as to decode and encode appendices, and then possible changes are required in EPP depending on how this is representated in the JSON.

This stylesheet is transforming an XML document by removing any "sec" element with a "sec-type" attribute value of "supplementary-material". These supplementary material sections are only partially retained by encoda - the files themselves need representation as downloadable files in the JSON. As it stands the labels and filepaths are decoded and re-encoded as paragraphs. The work to fix this is captured in https://github.com/elifesciences/enhanced-preprints-issues/issues/116. The current rendering on EPP as a result is less than ideal, where labels and filepaths are rendered in a series of paragraphs. The purpose of this xsl is to remove these sections until encoda and EPP can be updated to render downloadable supplementary files on the page.

Currently EPP client has no support for HTML tables, and no support for either mathML or latex representations of maths. Encoda will always use the machine readable version of content when available, i.e. the representation that is not an image. Therefore, when XML which has both machine readable and image representations of the same content, the image representation is stripped making the content unavailable in EPP. This xsl removes the machine readable representation of content when an image representation is available (both will be captured in an <alternatives> tag), so that the image version is at least rendered by EPP.

This xsl can be removed once support for machine readable tables (html) and maths (mathML and/or latex) is added in EPP and/or encoda is updated so that both representations are included in the JSON. Requirement captured in this ticket.

In JATS glossaries are captured using <glossary>. These typically contain a <def-list> (very similar to the description list in HTML). Encoda needs adjusting so as to appropriately decode the glossary (and def-list if it doesn't already) and encode this in JSON. After that EPP may need updating so as to render these appropriately.

In the meantime, this xsl transforms a def list into a simple list, with each definition/term pair being a list item in that list with their own p tags.

In JATS the <aff> element can be treated as mixed content. In other words, it can be somewhat prescriptive about how the content inside should be rendered - both the elements and text should be included in the order provided. In this ticket encoda was adjusted so that "any text not within the <institution> or address tags is appended to the organization name.". The issue with this approach alongside the approach on EPP is that the order of semantic content within aff is assumed - it's assumed that country will always appear last. This is not always the case, as is true for this preprint where the postcode appears after the country. In the JSON the country is pulled out of context, leaving extra punctuation in the name, and on EPP the affiliation is rendered with extra punctuation.

The same is true for other locale information within JATS (addr-line, city, state, and so on).

This can be solved in mutliple ways. One is to ask bioRxiv to not treat their affiliations as mixed-content - which has already been done. Another is for encoda to be adjusted so as to retain the original order of the content as it is supplied (I'm unsure if this will be possible).

In the meantime, this xsl simply strips all address/locale tags (retaining the content within) from any affs, so that the affiliation string can be rendered appropriately. It does mean that some semantic information is lost (the address in the HTML).

In JATS the <aff> element can be treated as mixed content. In other words, it can be somewhat prescriptive about how the content inside should be rendered - both the elements and text should be included in the order provided. In this ticket encoda was adjusted so that "any text not within the <institution> or address tags is appended to the organization name." When text content is present before an <institution> element inside <aff>, this content is appended to the <institution>, and the punctuation and content of the affiliaiton is as a result broken. For example

<aff id="a5"><label>5</label>Cell Biology Program, <institution>Sloan Kettering Institute, Memorial Sloan Kettering Cancer Center</institution>, New York, NY, USA</aff>

Gets decoded into:

{
  "type": "Organization",
  "name": "Sloan Kettering Institute, Memorial Sloan Kettering Cancer CenterCell Biology Program, ,, New York, NY, USA"
}

Sloan Kettering Institute, Memorial Sloan Kettering Cancer CenterCell Biology Program, ,, New York, NY, USA is not an acceptable rendering of this affiliation. As a result this xsl simply removes all <institution> elements from within aff, to skip this behaviour, and render the affiliation as originally provided.

It also strips any institution-ids (introduced as part of the Production process) from affiliations, as these are not intended for display.

This can be solved in mutliple ways. One is to ask bioRxiv to not treat their affiliations as mixed-content - which has already been done. Another is for encoda to be adjusted so as to retain the original order of the content as it is supplied (I'm unsure if this will be possible).

As explained in this comment from Nokome, when a journal reference has an <issue> tag as well as a <volume> tag in the XML, this affects the representation of that content in th JSON produced by encoda - Periodical is nested under both PublicationVolume and PublicationIssue. Because the client curently does not handle this level of nesting, the result is both a missing issue number and journal title for a reference.

This xsl removes the <issue> tag from a reference when it also has a <volume> tag, resulting in the level of nesting currently assumed by client in the JSON. This means that the journal title will be rendered by EPP (although the issue will still missing until support for that is also added).

The <etal> tag in JATS is used to indicate the presence of many unnamed contributors. This element is used when authors have made use of et al. in their references (rather than listing out all contributors).

Encoda does not decode this element, and therefore et al. is not included in the JSON giving the misleading impression that any authors listed before the et al are the sole contributors to that reference.

This XSL replaces the <etal> element with a surname as a workaround, meaning that this content is represnted in the JSON, and as a result within the HTML rendered by EPP. Support needs adding to encoda (and potentially subsequently client depending on the representation in the JSON), in order for this content to be rendered without this xsl.

JATS has a specific tag for elocation id (which is a piece of reference information somewhat equivalent to page number in online only journals) - <elocation-id>. Some vendors will use <fpage> instead of <elocation-id> to capture these, which while semantically incorrect, should be supported.

Encoda currently only decodes an <fpage> (<lpage>) if is an integer (as an aside this is not ideal either - page numbers are not always integers, and may include letters or other characters by convention). There is also a workaround in encoda noting that fpage is sometimes used in place of elocation-id, but this function will only handle cases where it starts with a e - which is convention that eLife use, but is not consistent within journal publishing.

This xsl will convert any fpage found in a reference that does not also have an lpage or elocation-id to elocation-id, so that it can be decoded by encoda and rendered by EPP.

This xsl is adding a missing affiliation link for all authors when there is only one affiliation. When this link is missing no affiliations display for any authors in EPP because encoda relies on the link to make add the affiliation for any/all authors. This can be solved with a tagging change which has already been requested from bioRxiv.

jats xml can accommodate alternative versions of names (e.g. westernised vs non-westernised names). When used these are tagged using named-alternatives. Encoda does not have support for this tagging and strips all names from the resultant JSON. This xsl deliberately mistags the alternative names so that one of them is included in a suffix element (which is supported by Encoda and rendered on EPP) as a workaround.

jats xml uses the element disp-quote for display quotes. These are decoded by encoda into QuoteBlocks, e.g.:

{
    "type": "QuoteBlock",
    "content": [
      {
        "type": "Paragraph",
        "content": [
          "Z-score = (value",
          { "type": "Subscript", "content": ["P"] },
          " – mean value",
          { "type": "Subscript", "content": ["P1…Pn"] },
          ")/standard deviation",
          { "type": "Subscript", "content": ["P1…Pn"] },
          ","
        ]
      }
    ]
  },

EPP does not currently support this content type and as a result the content within is completely lost in the HTML. This xsl strips the disp-quote element and includes the contents of any child paragraph elements so that (some of) the content is retained in the HTML.

In JATS, "[if] the <label> element is used in a <list-item>, it overrides the @list-type and @prefix-word attributes [on the <list> element]". Neither encoda or EPP have adequately taken this into account, and bioRxiv rely on this convention, meaning that some lists have two types of label indicator. This XSL adds the list-type attribute value simple when a list has list-items with labels (it will not do so if there are any list-items without labels within the list), so that only the indicator from the label is rendered in the HTML.

This XSL changes <boxed-text> into a <sec> element because encoda and EPP do not have adeqwuate support for this content.

For conference proceeding references, the name of a conference is usually captured using the element <conf-name>. Encoda does not decode/encode this element, as such it is missing in the HTML rendered by EPP. Encoda also does not distinguish in the reference type (still encoded as Article). This XSL converts a conference reference into a journal reference (as best as possible) so that the details can be showin in the HTML.

This xsl accounts for 'extra' abstracts captured preprints such as graphical abstracts, impact statements and 'highlights' sections. These would most appropriately be captured as separate (additional) abstracts, but EPP/Encoda is unable to retain the content when captured this way, so this xsl moves these so that they are sections wihin the main asbtract.

This xsl accounts for permissions for objects within xml. Encoda will decode the <license-p> within the permissions for a figure (I've not checked other objects) and encode this as licenses.content in the JSON. EPP does not currently render this content. Therefore this XSL will convert any permissions statement for an object into a paragraph which is added onto the end of a caption.

This xsl accounts for the capture of <code> in XML. Encoda correctly decodes this as CodeBlock, but EPP client does not currently render that content. If the <code> is inline this XSL will change it to <monospace>, otherwise if it's a block of code it will be changed to <preformat> (which incidentally is inappropriately decoded as a paragraph by Encoda, but at least surfaces the content).

We need support for CodeBlock added to EPP client. And we need Encoda to properly decode <preformat>.

Modify bioRxiv XML in preparation for Encoda

Prerequisites:

  • docker
  • libxml2-utils (if running on host)

XSL files

To apply an xslt transform to all biorXiv XML place it in the ./src folder.

To apply an xslt transform to a specific manuscript place it in the ./src/[DOI-SUFFIX] folder.

To apply an xslt transform to a specific version of a manuscript place it in the ./src/[DOI-SUFFIX] and express in the xslt the query to only apply the changes to that version number.

Tests for XSL files

Each xsl file must have at least one accompanying test. It is recommended that the ./test/fixtures/kitchen-sink.xml express at least one example which could be successfully targetted by the global xsl files in ./src/*.xsl.

You must drop an XML file in test folder for each XSL file in a folder name that corresponds to the filename of the xsl.

For example, an expected result of the ./src/change-label-and-title-elements.xsl transform can be found in ./test/change-label-and-title-elements. The filenames of the expected XML are the same as in the ./test/fixtures folder.

Some examples:

  • ./test/change-label-and-title-elements/kitchen-sink.xml contains the expected XML of ./test/fixtures/kitchen-sink.xml that has gone through the ./src/change-label-and-title-elements.xsl transform.

  • ./test/all/kitchen-sink.xml contains the expected XML of ./test/fixtures/kitchen-sink.xml that has gone through all the transforms directly in the ./src folder.

  • ./test/2022.05.30.22275761/remove-supplementary-materials/2022.05.30.22275761.xml contains the expected XML of ./test/fixtures/2022.05.30.22275761/2022.05.30.22275761.xml that has gone through the ./src/2022.05.30.22275761/remove-supplementary-materials.xsl transform.

Build docker image

docker buildx build -t epp-biorxiv-xslt .

Apply transform to XML

cat test/fixtures/2022.05.30.22275761/2022.05.30.22275761.xml | docker run --rm -i epp-biorxiv-xslt

Output to a file:

cat test/fixtures/2022.05.30.22275761/2022.05.30.22275761.xml | docker run --rm -i epp-biorxiv-xslt > output.xml

Introduce logging:

touch session.log
cat test/fixtures/2022.05.30.22275761/2022.05.30.22275761.xml | docker run --rm -i -v "./session.log:/session.log" epp-biorxiv-xslt --log /session.log

Apply only a single xslt:

cat test/fixtures/kitchen-sink.xml | docker run --rm -i epp-biorxiv-xslt /app/src/change-label-and-title-elements.xsl

Process a folder of biorXiv XML

The structure of the xml within the source folder will be preserved in the destination folder.

./scripts/process-folder.sh /path/to/SOURCE_DIR /path/to/DEST_DIR

Run with logs:

./scripts/process-folder.sh /path/to/SOURCE_DIR /path/to/DEST_DIR --log ./process-folder.log

Run tests

./project_tests.sh

Run with logs:

./project_tests.sh --log ./project-tests.log

Run projects tests entirely within docker container

docker buildx build -t epp-biorxiv-xslt:test --target test .
docker run --rm epp-biorxiv-xslt:test

Build docker image (api)

docker buildx build -t epp-biorxiv-xslt:api --target api .

Run api from image

docker run -p 8080:80 epp-biorxiv-xslt:api

Test api

curl --location 'http://localhost:8080' \
--data '<root><child>content</child></root>'

Test api (with passthrough)

curl --location 'http://localhost:8080' \
-H 'X-Passthrough: true' \
--data '<root><child>content</child></root>'