Skip to content

v1.3.0

Choose a tag to compare

@github-actions github-actions released this 05 Aug 01:53

Added

  • Version-history measurement now escalates automatically to whatever actually works on a given
    list, and always measures every eligible file

    Some lists don't project SharePoint's bulk version-size field at all; a few report it while
    leaving it unpopulated, which previously produced a confident but wrong 0 B for the whole
    library. The scan now proves a bulk mechanism actually carries real data (checking known file
    sizes against it) before trusting it, escalating through an un-expanded bulk sweep, then a
    CAML/RenderListDataAsStream sweep, and only then a per-file measurement pass — reserved for
    files whose version number doesn't already prove they have nothing retained. There is no
    longer a "Quick" cap that left some files unmeasured by design; every file that can have
    retained versions is now measured, and the per-file pass runs fast enough for that to be
    practical (requests are properly coalesced instead of running a few at a time).

  • Version History Count total
    Alongside the existing Version History Size total, the Storage Report's summary tile and Excel
    Summary sheet now also show the total number of retained versions across the whole scan — the
    same free per-file count, just summed.

  • Stage-by-stage scan progress, with an ETA and an explicit "paused" state
    The Storage Report's progress display now names exactly what's happening at every point in a
    scan — reading items, checking how a list can report version history, sweeping it in bulk,
    validating the result, or measuring per file — each with its own live count and, once there's
    enough data to estimate from, a time-remaining estimate. When SharePoint is throttling requests,
    the status line says so explicitly with a countdown, instead of the scan appearing to freeze. A
    hover tooltip shows additional detail (requests made, files found vs. added to the report,
    failed/unmeasured counts) without crowding the main display.

  • Independent cross-check of the version-history total
    For each library, the version-history total measured per file is compared against an
    independent figure derived from SharePoint's own folder-level Storage Metrics. Diagnostic only
    (logged, not shown in the UI) — it exists to catch a systematic measurement error before it's
    trusted.

  • Home screen
    The web part now opens on a landing screen with three cards — Tree
    View
    , List View, and Storage Report — each with a screenshot and
    a short description of what it does. Clicking a card opens that tool;
    every tool now has a Home button to return. A site administrator can
    still configure the web part to skip Home and open directly on any one of
    the three (or on Home itself) via the property pane's Default view on
    open
    dropdown, which gained Home, Tree View, and List View as
    options alongside the existing Storage Report.

Changed

  • The Explorer is now reached as two separate entry points: Tree View and
    List View

    Previously the web part opened directly into a single "Explorer" screen
    with an internal Treemap/List tab. That screen is unchanged internally —
    the same tab still switches between Treemap and List without losing your
    place — but it's now reached via the Home screen's Tree View or List
    View
    card (whichever is clicked decides which tab it opens on), and the
    header no longer shows "Explorer" / "Storage Report" tab buttons, since
    navigation moved to Home.

  • More accurate folder/library scans on fast tenants, without sacrificing
    the throttling protections that made the fallback walk bounded in the
    first place

    The fallback measurement walk (used when SharePoint's StorageMetrics
    rollup is stale or unavailable) previously stopped at a folder-count
    budget derived from Concurrent API requests (scanConcurrency * 150,
    capped at 2400) — on a fast tenant this budget was often exhausted in a
    few seconds, well before the 45-second wall-clock deadline that was
    already meant to be the real limit, cutting a scan short and reporting an
    approximate "≥" result for no real reason. The wall-clock deadline is now
    the sole practical bound; the folder count is a much higher safety valve
    (20,000) that exists only to catch pathological/runaway cases.

  • Folder/file listings that hit SharePoint's own paging limit are no
    longer silently under-counted

    A folder or file listing beyond getJsonPaged's internal page-count
    safety valve (previously 50 pages / 250,000 items, now 400 pages / 2
    million items) was truncated with only a console warning — the resulting
    total was short by whatever fell past the last page, with no indication
    to the user. This is now surfaced the same way other incomplete
    measurements are: as sizeApproximate in Tree View / List View, and as a
    counted, detailed skipped-folder entry in the Storage Report.

  • Saved Storage Report history no longer trims itself, and its "Partial"
    badge is gone

    A saved report used to keep only its Stale/Very-stale rows once it
    exceeded 50,000 files, flagged with a "Partial" badge — a mitigation for
    loading every saved report's full listing just to render the history
    list. History storage is now split into metadata (always loaded) and file
    listings (loaded only when viewed or compared), so every saved report now
    keeps every row. The only remaining trim is genuine storage exhaustion —
    browser quota running out evicts the oldest report's listing (never its
    summary) to make room, shown as a "No file list" badge instead.
    Comparing two reports now also degrades per-field instead of all-or-
    nothing: size and file-count deltas always work (they only need
    summaries); new/resolved archival-candidate counts are shown only when
    both reports' listings are actually available.

  • $batch envelope throttling no longer amplifies into a burst of
    individual requests

    A 429 on a coalesced $batch request was treated the same as a
    structural failure (a rejected header, $batch unsupported) and fanned
    out into one request per member — answering "you're sending too much"
    with up to 50× more requests, against a tenant that had just asked to
    slow down. A throttled batch now retries the whole envelope through the
    same shared backoff every other request uses; per-member fan-out is
    reserved for failures where the envelope itself will never work.

  • Excel Summary sheet values are left-aligned
    Numeric values (file/stale counts) previously sat right-aligned next to
    left-aligned text values (sizes, dates) in the same column, producing a
    ragged edge. Every value on the sheet now aligns the same way.

Fixed

  • A rejected bulk version-history field could either do nothing or cost
    far more than necessary

    Two separate issues on lists where SharePoint won't project the bulk
    version-size field: sharding used to give up on concurrency entirely
    rather than risk losing version data, even though the field's absence is
    now handled correctly by escalation; and the per-file fallback ran at a
    worker count sized for the pre-$batch era (3 workers), so envelopes
    never filled and measured roughly 3.5 files/second regardless of the
    $batch coalescing meant to speed it up. Both are fixed — sharding is
    re-enabled once the escalation can safely take over, and the per-file
    pass now runs enough workers to keep $batch envelopes full.

  • The Storage Report's progress bar and time estimate could both be
    confidently wrong while work continued

    A stale item-count estimate that got exceeded mid-scan was clamped rather
    than treated as unusable, which pegged the progress bar at 100% and could
    show "~0s remaining" for many minutes on a long-running scan. Producers
    now report no denominator at all once their own estimate is known to be
    stale, which the display renders as an indeterminate bar and no time
    estimate, rather than a wrong one.

  • ReportHistoryService.add() could hang indefinitely on a storage
    failure

    A synchronous exception from the underlying IndexedDB write (e.g. a
    clone failure, or the transaction aborting on quota) was never caught,
    so the returned promise neither resolved nor rejected — the scan
    appeared to finish but silently never saved, with no warning shown.
    Failures are now always caught and reported, and running out of quota no
    longer fails the save outright: the oldest report's file listing is
    freed first and the save is retried once before giving up.