Skip to content

Build1 publisher3 min readPublished

Confluence space exports omit blog posts from PDF without raising an error

Confluence Cloud's space export leaves blog posts out of a PDF and page comments out of HTML, and the file still finishes without a warning. What lands in it also depends on the permissions of whoever clicked.

The Engineer · Build desk

Illustration accompanying Confluence space exports omit blog posts from PDF without raising an error

What happened

  • Confluence Cloud exports a whole space to PDF, CSV, HTML or XML from space settings, an administrator action covering either the whole space or a set of pages picked individually.
  • A PDF export of a space leaves out blog posts, so a team that keeps decisions or release notes as blog posts exports a file without that body of writing in it.
  • Page comments are currently not exported during an HTML export, according to a dev.to walkthrough of what the space export actually produces.
  • None of these omissions produce an error: the file arrives looking complete, and the gaps get noticed weeks later by whoever needed the missing part.
  • Only content visible to the person exporting is included, with one exception: CSV or XML run by a site admin exports all content.

Compiled by The EngineerSomething wrong?How this is made

Why it matters

  • exposure A team holding a space PDF and an HTML archive can have lost both its blog-post history and its comment threads while believing the space is preserved in two formats.
  • decision Using export as a migration or archive path now needs a test run against your own space while the source still exists, because what goes missing changes with the format.
  • constraint A space deleted after a successful export cannot be rebuilt from the file, because Team Calendars are in none of the export formats.
  • cost Getting numbers out of wiki tables stays a retyping job, priced in the post at about six hours a year for a single quarterly report, paid by whoever assembles it.

Only content visible to the person clicking export is included in the file [8]. A restricted page that a space admin cannot read does not appear in the export that admin produces. Mikhail Klimenko, writing on dev.to, wrote: "The same button in two pairs of hands produces two different files, which is a problem when the export is supposed to be a complete record for legal or compliance reasons" [9][20].

Drafts are not in the export at all. An export takes published content, so an edit somebody has been carrying for three weeks does not exist as far as the export is concerned, and nothing on screen points that out [6].

The post maps destinations to formats: XML into another Confluence site, HTML for an archive a human can open later, PDF for sending outside, and it costs nothing [15]. Word never appears in the space-level list, because in Cloud it is a page-level action [2]. Klimenko wrote that anyone who tells you to just export the space to Word is describing Data Center or wishful thinking [18]. The 50-image ceiling people repeat for Word export is documented for Confluence Data Center, where a system property can raise it; Cloud has no such property, and the Cloud documentation does not state that number. He advises testing your own pages before planning around it either way [16].

The CSV option wastes the most time, because the name promises table data. It dumps the space's content, pages, attachments and comments, and the post says Atlassian's documentation does not describe its column layout [10]. Klimenko wrote that "the native export does not do this, and the CSV name is a trap" [11]. Copying tables by hand holds up until a merged cell appears, and then the values after it shift into the wrong columns while still looking like perfectly good numbers [12].

His cost figure is a measurement of his own reporting job. Eleven tables spread across a space, about four minutes each when nothing is merged, closer to ten when something is, which he rounds to an hour and a half, four times a year [13]. Eleven clean tables at four minutes is 44 minutes; eleven merged ones at ten is 110 [1]. The ninety-minute estimate therefore assumes merged cells in most of them, and it transfers to your team only if a report pulls from a similar number of tables and somebody spends the extra time reconciling the wiki figure against finance. Klimenko discloses that he builds Confluence apps, including one that exports tables to Excel or CSV with merged cells expanded, and writes that it does not replace the space export [17].

What to watch

  • Whether Atlassian changes the HTML export behaviour the post describes as page comments "currently" not being exported.
  • Whether Cloud documentation ever states an image limit for page-level Word export, which the post says it does not.
  • Whether a native Cloud export appears that returns table data, removing the retyping the post prices at an hour and a half a quarter.
Loading claim ledger
Loading source directory links
Loading share composer
Loading topic controls
Loading related stories