Allowed whitespace in ODFF and OOXML are
U+0020 SPACE
U+0009 CHARACTER TABULATION
U+000A LINE FEED
U+000D CARRIAGE RETURN
Line feed and carriage return look a bit funny in the Function Wizard if
part of a function's argument but work. Once a formula is edited, CR are
converted to LF though, probably already in EditEngine, didn't
investigate.
Change-Id: I6278f6be48872e0710a3d74212db391dda249ed2
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/119635
Reviewed-by: Eike Rathke <erack@redhat.com>
Tested-by: Jenkins
revert...
commit f97dbac73f
Author: Caolán McNamara <caolanm@redhat.com>
Date: Wed Jun 26 21:00:02 2019 +0100
infinite focus changing in toc biblio page
which blocked calling focus-changed callbacks when grab_focus is called
explicitly analogous to how we block value-changed callbacks setting
values through code but don't block them when the value is changed by
user interaction.
In retrospect that was a poor choice, so revert that and subsequent
workarounds in favour of just not calling gtk_grab_focus if the widget
already has focus.
checked:
a) tdf#138427 focus set to wrong input box
b) tdf#138078 only call GetFocus if we gained focus from an unfocused state
c) tdf#137993 ensure the toplevel vcl::Window is activated
d) tdf#136941 call focus in handler explicitly
Change-Id: I411480e2d627aa9995fb41b0aa17e9fb6d34d73f
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/114524
Tested-by: Jenkins
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
and for the normal case process immediately. Use-case is the
bibliography editor, modified uncommitted entry, click in browser
column margin area to select a new row, the entry should commit its
old contents to the old row before filling from the new row
Change-Id: Ib41d96afcfa86bcd1075b9512d4cfab593afa66d
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/111152
Tested-by: Jenkins
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
shave 5% of the time here - ref-counting triggered by copying
svl::SharedString is significant, so return by const&
instead of by value
Change-Id: Ic702632da45d75dddab33d6ce1e6f1097ff70de9
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/109900
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
in certain matrix formulas like SUM(IF(A1=G:G, H:H)*B1/B2) where whole
columns are used for comparison in the condition of IF ultimately
followed by a reducer like SUM. In such cases we can safely reduce the
double-refs involved in the comparison to the sheet area where there is
data before converting the data to ScMatrix.
This is a more restricted version of Noel's fix in
37ffe509ef011357123642577c04ff296d59ce68
Change-Id: I1c2e8985adedb3f4c4648f541fb0e8e7d0fae033
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/109050
Reviewed-by: Luboš Luňák <l.lunak@collabora.com>
Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com>
(cherry picked from commit 65167a9265acfea04733b5ff6ee3220a9da624f4)
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/109118
Tested-by: Jenkins
Reviewed-by: Dennis Francis <dennis.francis@collabora.com>
This should always have had a ParamClass member variable and
SetInForceArray()/GetInForceArray() functions but never did.
Now with
commit 3a33828b8d
CommitDate: Tue Oct 13 21:36:31 2020 +0200
Resolves: tdf#133260 Propagate ForceArrayReturn from inline arrays
it broke CppunitTest_sc_subsequent_filters_test with the virtual
dummy assert() in FormulaToken::SetInForceArray(). Unfortunately
not caught earlier but at least it did now.
Remove the duplicated FormulaToken overrides and simply derive
from FormulaByteToken instead that has all we need.
Change-Id: I7946602955970fb9d6f349efdacb41389e211b68
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/104262
Reviewed-by: Eike Rathke <erack@redhat.com>
Tested-by: Jenkins
Recheck after 7-0 branchoff
Found with bin/find-unneeded-includes
Only removal proposals are dealt with here.
Change-Id: I7f5c35474b671facbda85a9538462dd4291add91
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/97655
Tested-by: Jenkins
Reviewed-by: Miklos Vajna <vmiklos@collabora.com>
(*) fix: I was substracting the padding space instead of adding it
when calculating how much free space we had to improve.
(*) sort input data, so we process structs located in the same DSO
together, which reduces GDB's memory usage
(*) handle another error condition, where gdbs output is sufficiently
mixed up that we miss the end of commands terminator
Change-Id: Ic4bb92b736f38a2b3d90e4a14485152b7f869b43
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/95041
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
Same as RAND() and RANDBETWEEN() but not recalculating on every
change, just the normal expression recalculation.
Change-Id: I8ba7099125e487a78bd3d91db8b666c2f36b22fd
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/92994
Tested-by: Jenkins
Reviewed-by: Eike Rathke <erack@redhat.com>
and reuse them for interpret'ing all cells under the
respective threads. This gives a sizeable win in the execution
time especially for long formula-groups.
Change-Id: Ib340950f21e863b5b821d20c092214d8bc5012aa
Reviewed-on: https://gerrit.libreoffice.org/80845
Tested-by: Jenkins
Reviewed-by: Luboš Luňák <l.lunak@collabora.com>
Found with bin/find-unneeded-includes
Only removal proposals are dealt with here.
Change-Id: I564e3dd53e7718326fc1e8b98c9ae3367381703c
Reviewed-on: https://gerrit.libreoffice.org/70550
Tested-by: Jenkins
Reviewed-by: Miklos Vajna <vmiklos@collabora.com>
this is first of the calc dialogs with a range selector, so some temp
scaffolding is required during interim case of both welded/unwelded
in existence
Change-Id: I5480179092da7b56864cef066af781b35f735ebc
Reviewed-on: https://gerrit.libreoffice.org/70474
Tested-by: Jenkins
Reviewed-by: Caolán McNamara <caolanm@redhat.com>
Tested-by: Caolán McNamara <caolanm@redhat.com>
FOURIER(Array, GroupedByColumns, Inverse, Polar)
is a matrix formula that computes discrete Fourier transform[DFT]
of input array(first argument) via a radix-2, decimation-in-time
fast Fourier transform algorithm.
Unit test for this is coming up in a new commit.
The data in input array(first argument) can be :-
1) grouped by columns (needs to be indicated by flag GroupedByColumns = TRUE)
In this case the array can contain 1 or 2 columns, where the first column
contains the real part of input series and second column if present contains
the imaginary part of the input series. If there is only 1 column, the input
series is treated as purely real. If the number of rows is not a power of 2,
zeroes are appended to the input series internally to make the series length
equal to the next nearest power of 2.
2) grouped by rows (needs to be indicated by flag GroupedByColumns = FALSE)
In this case the array can contain 1 or 2 rows, where the first row
contains the real part of input series and second row if present contains
the imaginary part of the input series. If there is only 1 row, the input
series is treated as purely real. If the number of columns is not a power of 2,
zeroes are appended to the input series internally to make the series length
equal to the next nearest power of 2.
The third argument "Inverse" is a boolean flag to indicate whether an inverse
DFT needs to be computed. This argument is optional and the default value is
FALSE.
The fourth argument Polar is a boolean flag to indicate whether the final output
needs to be in polar coordinates. This argument is optional and the default value
is FALSE.
The result of DFT consists of two columns - first column contains the real parts (or
the magnitudes if Polar=TRUE) and second column contains the imaginary parts (or
the phases if Polar=TRUE).
Implementation:
A fairly straighforward non-recursive implementation of radix-2 FFT algorithm is
written from scratch.
Reference:
Heckbert, P., 1995. Fourier transforms and the fast Fourier transform (FFT) algorithm.
Computer Graphics, 2, pp.15-463.
The normalization factor used in DFT / and inverse DFT in this implementation matches that
of fft() and ifft() functions of Matlab/Octave. It also matches the one used in Wikipedia
article on DFT: https://en.wikipedia.org/wiki/Discrete_Fourier_transform.
Change-Id: If4a40a6ef62bce1f03c589ae5357b2049f66fe64
Reviewed-on: https://gerrit.libreoffice.org/67938
Tested-by: Jenkins
Reviewed-by: Eike Rathke <erack@redhat.com>
FREQUENCY() forces its direct caller into array mode, but only for
the immediate subexpression and not for further operators of the
same parameter.
This weird Excel behaviour is stated in
ECMA-376-1:2016 OOXML 18.17.7.127 FREQUENCY
"A call to FREQUENCY shall be an array formula."
somewhat unclear what it actually applies to, but it turned out
that "a call" is indeed *only* THE direct call, see
https://bugs.documentfoundation.org/show_bug.cgi?id=122301#c19
Change-Id: I145d8fe26d75d5af25b987e190bf35f2d2c03ec6
Reviewed-on: https://gerrit.libreoffice.org/66407
Reviewed-by: Eike Rathke <erack@redhat.com>
Tested-by: Jenkins
Too many side conditions are possible with implicit array of
references in array mode. Propagate new ReferenceOrRefArray
parameter class to indicate the preferred return type for
arguments to functions whose parameters explicitly handle array of
references.
Change-Id: I1f01266495c2ef1941ffe0cb7c2e0a5ae0bb7e69
Reviewed-on: https://gerrit.libreoffice.org/63201
Reviewed-by: Eike Rathke <erack@redhat.com>
Tested-by: Jenkins
Since ScTokenArray::Add() overallocates memory, make sure we do not
keep such possibly large arrays. Since any copying of ScTokenArray
implicitly finalizes as well, this is not a big problem right now,
but then why needlessly do the copies? (next commit)
Change-Id: I55398bcd8fb31f1be5a4b8e3f5a71b26649a7594
Reviewed-on: https://gerrit.libreoffice.org/60862
Tested-by: Jenkins
Reviewed-by: Luboš Luňák <l.lunak@collabora.com>
Eike Rathke