FOP-2469: Auto table layout

[chkpnt]

Draft for FOP-2469 by applying FOP-2469-2018-07-17.patch.

[chkpnt]

What could be the reason for commenting out this assertion? I've compared the patches in FOP-2469 and this was introduced in Eric Lim's version of the patch. @esclim Do you remember, why you did this 7 years ago? ;-)

[chkpnt]

Raw type

[esclim]

Sorry I don't remember. Way too long ago.

…

-Eric

On Thu, Oct 30, 2025 at 7:17 AM Gregor Dschung ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In fop-core/src/main/java/org/apache/fop/layoutmgr/BreakingAlgorithm.java <#95 (comment)> : > @@ -1405,7 +1405,7 @@ protected KnuthNode getNode(int line) { * @return the width/length in millipoints */ protected int getLineWidth(int line) { - assert lineWidth >= 0; +// assert lineWidth >= 0; What could be the reason for commenting out this assertion? I've compared the patches in FOP-2469 <https://issues.apache.org/jira/browse/FOP-2469> and this was introduced in Eric Lim's version of the patch. @esclim <https://github.com/esclim> Do you remember, why you did this 7 years ago? ;-) — Reply to this email directly, view it on GitHub <#95 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABIV4GQO7NLDULVOBBNEIM332HXTXAVCNFSM6AAAAACKAPUS5GVHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZTGNZQGM2TQMJVGU> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[siebej]

While stress-testing this patch on a 943-page document (Dutch government publication with 536 auto-layout tables) we hit an NPE in ColumnSetup.redistributeAuto: tLM.getPossibleWidths(tcol, context) returns null for columns without PrimaryGridUnits (empty columns in a table with a fixed width), but unlike redistribute() this method does not null-check before calling getMax().

Suggested fix, mirroring the null handling in redistribute():

MinOptMax possibleWidth = tLM.getPossibleWidths(tcol, context);
if (possibleWidth == null) {
    // no PrimaryGridUnits in this column; its width is zero
    continue;
}

With that fix (and a rebuilt event model for columnsInAutoTableTooWide) the whole 943-page document renders in ~9 seconds with sensible content-based column widths. Happy to share test material if useful.