Skip to content

Commit

Permalink
v1.4 optimizations
Browse files Browse the repository at this point in the history 
Speeds up flattening of differentially-kerned text, which is now the default for PDF imports
Adds a second all differentially-kerned text test, to be used for adjusting remove_kerning (merge logic will need to be tweaked for the new kernings)
  • Loading branch information
@burghoff
burghoff committed Oct 18, 2024
1 parent e242d7d commit cde3140
Show file tree
Hide file tree
Showing 5 changed files with 2,957 additions and 74 deletions.
159 changes: 87 additions & 72 deletions scientific_inkscape/inkex1_3_0/inkex/text/parser.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -318,6 +318,8 @@ def duplicate(self):
ret_ln = line.copy(cmemo)
ret_ln.ptxt = ret
ret.lns.append(ret_ln)

ret.affected_cnts = {cmemo[k]: [cmemo[v2] for v2 in v if v2 in cmemo] for k,v in self.affected_cnts.items()}

# nextw et al. could be from any line, update after copying
for ret_ln in ret.lns:
Expand Down Expand Up @@ -359,7 +361,8 @@ def parse_lines(self, srcsonly=False):
# Next we find the top-level sodipodi:role lines
xvs = [ParsedText.get_xy(ddv, "x") for ddv in dds]
yvs = [ParsedText.get_xy(ddv, "y") for ddv in dds]
nspr = [ddv.get("sodipodi:role") for ddv in dds]
sprs = {ddv:ddv.get("sodipodi:role") for ddv in dds}
nspr = [spr for spr in sprs.values()]
nspr[0] = None
sprl = [
nspr[i] == "line" and len(xvs[i]) == 1 and len(yvs[i]) == 1
Expand Down Expand Up @@ -452,8 +455,18 @@ def inherit_none(iin, xyt):
# Finally, walk the text tree generating lines
lns = []
sprl_inherits = None
deltacnt = 0;
affected_cnts = dict()
for ddi, typ, tel, sel, txt in self.tree.dgenerator():
newsprl = typ == TYP_TEXT and types[ddi] == "tlvlsprl"
if (
typ == TYP_TAIL
and sprs.get(tel) == "line"
and self.tree.pdict[tel] == self.textel
):
deltacnt += 1
affected_cnts[tel] = []

if (txt is not None and len(txt) > 0) or newsprl:
sty = sel.cspecified_style
ctf = sel.ccomposed_transform
Expand Down Expand Up @@ -576,7 +589,7 @@ def inherit_none(iin, xyt):
for j, c in enumerate(txt):
csty = self.font_picker(txt, j, fsty, tsty)
prop = self.ctable.get_prop(c, csty)
TChar(
tc = TChar(
c,
fsz,
scf,
Expand All @@ -585,15 +598,21 @@ def inherit_none(iin, xyt):
sty,
csty,
CLoc(tel, typ, j),
lns[-1],
lns[-1], deltacnt
)
for afd in affected_cnts:
# inkex.utils.debug(c)
affected_cnts[afd].append(tc)
deltacnt += 1

if j == 0:
lsp0 = lns[-1].chrs[-1].lsp
bshft0 = lns[-1].chrs[-1].bshft
else:
lns[-1].chrs[-1].lsp = lsp0
lns[-1].chrs[-1].bshft = bshft0

self.affected_cnts = affected_cnts # so src-only calls don't change
return lns

def finish_lines(self):
Expand Down Expand Up @@ -742,21 +761,24 @@ def get_delta_num(self):
to the top-level text"""
allcs = [c for line in self.lns for c in line.chrs]
topcnt = 0
for _, typ, ddv, _, txt in self.tree.dgenerator():
for _, typ, src, _, txt in self.tree.dgenerator():
if (
typ == TYP_TAIL
and ddv.get("sodipodi:role") == "line"
and ddv.tag == TStag
and self.tree.pdict[ddv].tag == TEtag
and src.get("sodipodi:role") == "line"
and src.tag == TStag
and self.tree.pdict[src].tag == TEtag
):
topcnt += 1
# top-level Tspans have an implicit CR at the beginning of the tail
if txt is not None:
for i, _ in enumerate(txt):
thec = next((c for c in allcs if c.loc == CLoc(ddv, typ, i)), None)
thec = next((c for c in allcs if c.loc == CLoc(src, typ, i)), None)
if thec.topcnt != topcnt:
inkex.utils.debug((thec.line.ptxt.textel.get_id(),thec.c,thec.topcnt-topcnt))
thec.deltanum = topcnt
topcnt += 1


def get_delta(self, xyt):
"""Calculates the delta values for the text lines."""
for ddv0 in self.tree.dds:
Expand All @@ -766,20 +788,19 @@ def get_delta(self, xyt):
if len(dxy) > 0 and dxy[0] is not None and not (self.isflow):
allcs = [c for line in self.lns for c in line.chrs]
cnt = 0
for _, typ, ddv, _, txt in self.tree.dgenerator(subel=ddv0):
for _, typ, src, _, txt in self.tree.dgenerator(subel=ddv0):
if (
typ == TYP_TAIL
and ddv.get("sodipodi:role") == "line"
and ddv.tag == TStag
and self.tree.pdict[ddv].tag == TEtag
and src.get("sodipodi:role") == "line"
and self.tree.pdict[src] == self.textel
):
cnt += 1
# top-level Tspans have an implicit CR at
# the beginning of the tail
if txt is not None:
for i, _ in enumerate(txt):
thec = next(
(c for c in allcs if c.loc == CLoc(ddv, typ, i)), None
(c for c in allcs if c.loc == CLoc(src, typ, i)), None
)
if cnt < len(dxy):
if xyt == "dx":
Expand All @@ -804,14 +825,16 @@ def update_delta(self, forceupdate=False):
# only if new is not old and at least one is non-zero

if anynewx or anynewy or forceupdate:
self.get_delta_num()
# self.get_delta_num()
# for c in self.chrs:
# c.deltanum = c.topcnt
dx = []
dy = []
for line in self.lns:
for c in line.chrs:
if c.deltanum is not None:
dx = extendind(dx, c.deltanum, c.dx, 0)
dy = extendind(dy, c.deltanum, c.dy, 0)
if c.topcnt is not None:
dx = extendind(dx, c.topcnt, c.dx, 0)
dy = extendind(dy, c.topcnt, c.dy, 0)

if not (self.flatdelta): # flatten onto textel (only do once)
for ddv in self.tree.dds:
Expand Down Expand Up @@ -934,50 +957,7 @@ def make_editable(self):

def split_off_chunks(self, chks):
"""Splits off specified chunks of text into a new element."""
nll = self.duplicate()
newtxt = nll.textel

iln = self.lns.index(chks[0].line)
# chunks' line index
wiis = [chk.line.chks.index(chk) for chk in chks] # indexs of chunks in line

# Record position and d
dxl = [c.dx for chk in chks for c in chk.chrs]
dyl = [c.dy for chk in chks for c in chk.chrs]

for chk in reversed(chks):
chk.del_chk()

# If line was continuing, fuse the coordinates
if nll.lns[iln].continuex:
nll.lns[iln].continuex = False
nll.lns[iln].xsrc = nll.lns[iln].chrs[0].loc.sel
nll.lns[iln].change_pos(newx=nll.lns[iln].x)
if nll.lns[iln].continuey:
nll.lns[iln].continuey = False
nll.lns[iln].ysrc = nll.lns[iln].chrs[0].loc.sel
nll.lns[iln].change_pos(newy=nll.lns[iln].y)

# Delete the other lines/chunks in the copy
for il2 in reversed(range(len(nll.lns))):
if il2 != iln:
nll.lns[il2].dell()
else:
nln = nll.lns[il2]
for j in reversed(range(len(nln.chks))):
if not (j in wiis):
nln.chks[j].del_chk()

cnt = 0
for ln2 in nll.lns:
for c in ln2.chrs:
c.dx = dxl[cnt]
c.dy = dyl[cnt]
cnt += 1
nll.update_delta()

newtxt._parsed_text = nll
return newtxt
return self.split_off_characters([c for chk in chks for c in chk.chrs])

def split_off_characters(self, chrs):
"""Splits off specified characters into a new element."""
Expand All @@ -1001,7 +981,7 @@ def split_off_characters(self, chrs):
oldy = chrs[0].chk.y + dyl[0]

for c in reversed(chrs):
c.delc()
c.delc(updatedelta=False)

# Delete the other lines/chars in the copy
for il2 in reversed(range(len(nll.lns))):
Expand All @@ -1011,7 +991,7 @@ def split_off_characters(self, chrs):
nln = nll.lns[il2]
for j in reversed(range(len(nln.chrs))):
if j not in ciis:
nln.chrs[j].delc()
nln.chrs[j].delc(updatedelta=False)

# Deletion of text can cause the srcs to be wrong.
# Reparse to find whereit is now
Expand Down Expand Up @@ -1039,6 +1019,7 @@ def split_off_characters(self, chrs):
c.dy = dyl[cnt]
cnt += 1
nll.update_delta()
self.update_delta()

newtxt._parsed_text = nll
return newtxt
Expand Down Expand Up @@ -1359,6 +1340,7 @@ def parse_lines_flow(self):
For non-rectangular flow regions, uses svgpathtools to find where
lines are drawn (this is imported conditionally).
"""
self.affected_cnts = dict();
sty = self.textel.cspecified_style
isflowroot = self.textel.tag == FRtag
isshapeins = (
Expand Down Expand Up @@ -2188,6 +2170,13 @@ def dell(self):
c.loc.elem.tail = (
c.loc.elem.tail[: c.loc.ind] + c.loc.elem.tail[c.loc.ind + 1 :]
)

myln = self.ptxt.lns.index(self)
for ln in self.ptxt.lns[myln+1:]:
for c in ln.chrs:
c.topcnt = c.topcnt - len(self.chrs) - self.sprl if c.topcnt is not None else None
# inkex.utils.debug('updating' + c.c + 'in '+self.ptxt.textel.get_id())

self.ptxt.update_delta()

if self in self.ptxt.lns:
Expand Down Expand Up @@ -2242,6 +2231,11 @@ def disablesodipodi(self, force=False):
newsrc = self.chrs[0].loc.elem # disabling snaps src to first char
if self.chrs[0].loc.typ == TYP_TAIL:
newsrc = self.chrs[0].loc.elem.getparent()
if newsrc.get("sodipodi:role")=='line':
if newsrc in self.ptxt.affected_cnts:
for c in self.ptxt.affected_cnts[newsrc]:
c.topcnt -= 1;
del self.ptxt.affected_cnts[newsrc]
newsrc.set("sodipodi:role", None)

self.sprlabove = []
Expand Down Expand Up @@ -2485,10 +2479,10 @@ def y(self):
return lny[i] if i < len(lny) else lny[-1]
return 0

def del_chk(self):
def del_chk(self,updatedelta=True):
"""Deletes the chunk from the line."""
for c in reversed(self.chrs):
c.delc()
c.delc(updatedelta=updatedelta)
if self in self.line.chks:
self.line.chks.remove(self)

Expand Down Expand Up @@ -2553,6 +2547,17 @@ def appendc(self, ncv, ncprop, ndx, ndy, totail=None):
newx = self.line.x
newx[self.chrs[0].lnindex] -= deltax
self.line.change_pos(newx)


# update topcnt
self.line.chrs[myi].topcnt = self.line.chrs[myi-1].topcnt+1
for c in self.line.chrs[myi+1:]:
c.topcnt += 1
myln = self.line.ptxt.lns.index(self.line)
for ln in self.line.ptxt.lns[myln+1:]:
for c in ln.chrs:
c.topcnt += 1

self.line.ptxt.update_delta()

def append_chk(self, nchk, stype, maxspaces=None):
Expand Down Expand Up @@ -2950,7 +2955,7 @@ def style_derived(styv, pel, textel):
class TChar:
"""Represents a single character and its style."""

def __init__(self, c, tfs, scf, utfs, prop, sty, tsty, loc, line):
def __init__(self, c, tfs, scf, utfs, prop, sty, tsty, loc, line, tcnt=None):
"""Initializes TChar with given parameters."""
self.c = c
self.tfs = tfs
Expand Down Expand Up @@ -2988,8 +2993,6 @@ def __init__(self, c, tfs, scf, utfs, prop, sty, tsty, loc, line):
# get later
self._dy = 0
# get later
self.deltanum = None
# get later
self.dkerns = lambda cL, cR: prop.dkerns.get((cL, cR), 0) * utfs
self.parsed_pts_t = None
self.parsed_pts_ut = None
Expand All @@ -2999,6 +3002,7 @@ def __init__(self, c, tfs, scf, utfs, prop, sty, tsty, loc, line):
# letter spacing
self.lhs = None
# flow line-heights
self.topcnt = tcnt

def copy(self, memo=None):
"""Creates a copy of the TChar instance."""
Expand Down Expand Up @@ -3152,8 +3156,11 @@ def get_baseline(styin, fsel):
else:
bshft = ipx(bshft) or 0
return bshft

def __str__(self):
return str((self.c,self.loc.elem.get_id(), self.loc.typ, self.loc.ind))

def delc(self):
def delc(self,updatedelta=True):
"""Deletes the character from the document and its chunk/line."""
# Deleting a character causes the chunk to move if it's center- or
# right-justified. Adjust position to fix
Expand Down Expand Up @@ -3216,6 +3223,14 @@ def delc(self):
changedx = True
if changedx:
self.line.change_pos(lnx)

# Update topcnt
for c in self.line.chrs[self.lnindex+1:]:
c.topcnt -= 1
myln = self.line.ptxt.lns.index(self.line)
for ln in self.line.ptxt.lns[myln+1:]:
for c in ln.chrs:
c.topcnt -= 1

# Delete from line
_ = [
Expand All @@ -3238,9 +3253,10 @@ def delc(self):

# Remove from chunk
self.chk.removec(self)

# Update the dx/dy value in the ParsedText
self.line.ptxt.update_delta()
if updatedelta:
self.line.ptxt.update_delta()

def add_style(self, sty, setdefault=True, newfs=False):
"""Adds a style to the character by wrapping it in a new Tspan."""
Expand All @@ -3254,7 +3270,6 @@ def add_style(self, sty, setdefault=True, newfs=False):
tafter = prt.text[self.loc.ind + 1 :]
prt.text = tbefore
prt.insert(0, t)

t.tail = tafter
else:
tbefore = prt.tail[0 : self.loc.ind]
Expand Down
2 changes: 1 addition & 1 deletion tests/data/refs/flatten_plots__--id__layer1__--testmode__True__Text_tests__svg.out
View file

Large diffs are not rendered by default.

1 change: 1 addition & 0 deletions tests/data/refs/flatten_plots__--id__layer1__--testmode__True__Text_tests_dx__svg.out
View file

Large diffs are not rendered by default.

Loading

0 comments on commit cde3140

Please sign in to comment.