From c30d69d205c88088253d6917e1fa96113c370bf4 Mon Sep 17 00:00:00 2001
From: Marcel Gansfusz <marcel.gansfusz@gmail.com>
Date: Thu, 23 Oct 2025 00:06:25 +0200
Subject: [PATCH] added back option to run OCR

---
 app/__pycache__/main.cpython-313.pyc | Bin 33234 -> 35835 bytes
 app/main.py                          |  93 +++++++++++++++++++++++----
 index.html                           |   5 +-
 static/app.js                        |   4 +-
 4 files changed, 84 insertions(+), 18 deletions(-)

diff --git a/app/__pycache__/main.cpython-313.pyc b/app/__pycache__/main.cpython-313.pyc
index dab1333e555aa09c5b5ab1ade37bfede83d66d3b..098bd0564940a80d154e7ba45b66adaf6ec2b140 100644
GIT binary patch
delta 3746
zcmaJ@32;-#8Qy);(=FMuB_AS=o_t{BOU6e)fDFbqBBj{iXMr6HWFgDO*p|$bY#;}A
zN;9F!BoW<a0wtZ$bfyh6bV^3;G|5atI}J&iV~P_Rt7_8FX;Ye}hbYV>G%eHrf3gT6
z$ByQE|9=0myZ`>Xt2YmEpFGByE*p(H0?i^`-Q7HQ#8g6`e7bs@+SuxMx9G$>Y-Ep8
z+QPo3Zm!BQgEF&K^!q(xgTJ>$DejONQQYY->V1lB-BxJN@&TFmi9>!jShj&>P-eEV
zl-fhLvm#A>S-<F&wMxGS#^ZRH2-t;Xy8$+&vC}))oTi!%vZplmS-<FkQQdwI*o@0I
zJH!cpFU~_$*Q;gUNtxd%M&!6|u#Cz~K)grhrXYfB3tsLPmADtRd#E%dcKA8J31=W{
zl%gd21;rRdm!lbEhqz4dVi{#I8`Bo$g~WY0b#cGU4<{ei)>kSN(GdG(eKALr$#ne>
zIh{{03ak1F2`*y?8(Yg1tt!kpA-5CJzFR@a41HA5%^Qe7daBu6qr)o5Y5-rp&+Ie!
zj6U#TuQXPaI(!Nf5zBpM6KPhq(yX081}m7O(vm+|b=}b{R*02s+47)%Rztc6RwzcO
zvx+5_@8{KZLNt`jZ~8jrm3`85m<>0tq=oEz&CxtlC>)7JrNGqiNFW-Ln7-wV?kt3S
z7zT=c$j-OywuZS3AtT%ve!|=y{AkD7*4EK4)+FC(T}lHfweIl9p?lsjSr<(9H%<0;
z47QZPk}~9{427~#aM`HN(|ttLTHPgtD+MI4C}&&rrM4N_*2ZdA)N$>U^{-gDBOgN)
z09=jrz|6nVajUnD{;IMO@I|Z1Ya}mLv=>9?>?-PAr98X38swMCsdtU?rHUGmUpDHz
zD-<smFYz`hUT)%1UZF&O4fVDq-(FF{@vlPM_mkJ$OE}xFL5$%p2$C02dL;)}So4}w
zT!?2Mtm)ds`*@LJZADZ*zE=stO)a7xg!;`kpE|d#g~UrY8o*1_-@oXsVFx^A)yuHw
zBCGa6N&SS7LH%`KP0m*!J{5c3<1Ac>wxWEt>cK9e1^%kXY0vTs)XKcBUyfQxHf{kc
zJ!@}s%<8kV_G#<E42oEZ);`^0PhJtQyc`2HpS9DAs_Porr*(1V5ra?FYXyg!n$IAr
zMU9&i$mU)`^h7ByqP@z()`c|<(9wv>dhi)6$vOJu8Hw6zAbjc@dFEPMT%6S!Pzx8=
zV)H54j<xlaXAiD*0RDKb>nrTalh$>YC_Ty4>#OPO%(ecM!zkSck)(47ClTfm;B*t|
zy9f)6cG%qSftco}q;M!6GlDGK*s`_q0$L*w4vOeIP(^d(N_o=P@hB|`VdioTZ)mtB
z$mLUtZ}(Ojd?dHY5uUaA1y)<7H9}!6mYCYdv(+A#x(_$Ie}L`vILuEG((lBOi$!yS
z{lIgKZe^afW3!lY1*BZWltYrWi%BlZ4gZ-WYX)!Tb(mgJ|2aOJ_M5h<cS$a4?#q7R
z*WpI1S#YhzAR0xJn?_7NQ?DH#6Jws3FPhz4Bo9i)!6w3Gw44R2EVI5mvA|7piGU06
z`WX-lb4!4W(&EzqTATS#ExmN}_Qj>DdaH8N`1I?$A*xvPCCf5@Ze8rO-!>qS(PC#w
zkb@2FkZ;nM09UCHj0_9XbR;J9boA=_c87(Lv5BxS7863z$V^y@hlhoDRM;}MKRgkW
zjpMt+5dmsP3I{>ipc9j>b<fgS=`EP1bRMAwp%$SI;kN)zF3pFc6H!Te8>t0^6u<`Q
zbrkOcI7!or;i)lz@brHyuIZTMitP?cVOKmHiwy_kLDxwxt(gc$MyG?L;j|`4gP$rE
zpM~j6-pOwPl^{X6kUR5N2_0SCVo&ew-Cm)yr@O1K+gT_*gu^|I@Gin15P}FlMZlAB
z!$K}VH?19(g8Kq^SR@y!8xa148r^6(9+(>2KN*}#&zc~a@GJ%5(Om~%YSJTUq3Mi<
zrYFOZxb!}Bo%vsVJkzS^bbM+$o=wkRaGZ+>e?|BJL7v~=KuW82jm2?XEk1`Zq6U(b
z4+?>8t+BAS4&Bm5Jh>OLCwD2)7d%q;KwTpKmhP6hj(3cfNBj@?AKW>&frUC=bv$M|
zVoF&`Q?`m!X?aFX%=XJfX)d^wM@k);64JPw4Ro%ZJ+EoHV75MHKVm;2ChX_T&cmvd
z#kOE^&Rd)*t1V@%&G5uplOYPL?yxqaA=<nJP3gR*^zrHiTg{ucn$ymC+nVzl&!08M
z1fK{VbsPyNwjXUj8a}%HTxH8SQ|p4}wo4vjvSr!`ufKoOy_*u6^UBgxt?Rt9_$Yg=
zQ=k{w-#c3X9bLDUUW4V1<0A3$m?J5@O@ikEFS$PsTi8U`j^)xnu>UWF48kP<CntS`
z^2aDEvZ?+AiT?sH_l6}rft*E?0~<C`8wK&kCv+SQ5$;aD;5ATAMcL~c>*#&Ss~b<K
zRPwcZkv-AlqAdBFo+p*`cTDPA!zm~`-PfthgJ2TEOUY|}9SW@x#55J|T*^)e7Phnh
z30lvt_1~t;M>~XflWhYVIbiP}Y|s@Vi|}#s%-}y2sv7JsW63}#jPh}yl^R*YP#bl!
z$)RU%S&qs^0B4@0K)Dv-S)6?xYY(n$XhNMyawEM8AgvnT7nDY0X?|iRI5sRb$lSzC
zBsdwC%Gi&B>-6<#i@+<A{|-LK>6?*2s8b|=K75Gs(iKc$=STxDT}A2cks#NsNWM4n
zeM<ewdw1W(Rd=IL2<NdGq!>QB`1oWWog>I2j3z6_0}5(oha#(KH#-xlrS~L1jJ(Iu
z&Fr<kOK2i_ac>?6`j28g{7wjTJh>_UGz`N2H{DcV#W8ju?1EUZ-vQP>W7VC(=m@9S
z(9HCbUNCU-(jFkXySpX1OZ%L(+Zkq=ndek;+N|vKz8YG>e!owmX4bvG*j@l`>4L8Q
zKxcbjSD*vH3pY%cC>7&?7WU2k4-KBhkzPf>zg%g4WFi`j%de^Qyi8<o2&~_<5^pRC
zOHQsUX#(AzL6FNut_6AP<@7hR(F0{FT(NY9J#e5*cMwGQkys%4^nuOF^&+M1G^H&_
zG#^tQS0z@@Ywa0@OkU6xCOVE;ju$01&Fh2=yvSggQ4xbRAs(wdzBDm7uXkkBB3pN`
zs3fb@5%9SokX~ew>q-+b79~Q*T8`hA7@s$kXYxe$_`%Xe&!U^c7K`kSgQcC>u(uMk
zEv3<atM9>WhxL~foLQBzXEl|d9q@9MOkY+It|-H0O(s~)Y?Ha3x>7o0MgiXoCUb_b
zXOpwVZs@_SV>Bl$iO|v3#GZMx181g1UPnykD^SC(%n)6<=nY@ldnI~TL+=Z{k<WUg
LG?V<<Y#H}|E)sg1

delta 2306
zcmZvcdrVu`9mmhPFJE6jsGWcT2QZHS&WK3}B#_RCgoo2m!sB8}zyYp}3Fcv(dm&+q
zHfYhLptXs+U#5k!s_mp@tJ*)B-A#RL%^zFRtee!1PU|c;>M~73nkrSUU`wn1vvaPK
zEZufwfA8;p&hLEB<Nogb@(0WZD~x^1W|I(UiOw%3KRx@bJs(e8s~RwQheFMAL#V0V
zDi0T;DNgN?Lm|KH3-#7>@+pJ#$s?h>zN_%k*pAXHA2oQtJQivqnaw0qF-TAj!#*(|
zd}28sfIe|Aegc+7Hy(r^iY53Y{7S4kH6%}xQ88mA7xI&t8J;YwUXZ5@k73zQ=5Gr5
ziR-jMLh?+gm##=Q?KP21PM$URDLHPqG!ZgqkP&Dxm*7ztGrL@4vPMSS&k*E!BSV1<
z=Yk)b^K!!SIXbtzXz+;sJM-QOuMmmGwK+9nr6+Z-<&o<O`EUJ<MW`Hk(D%_WhB;ra
z9`|)JaC6^*T#M{EJcy9OAS9P|W4SEIHkhDuf4RRr$l8%y5p>wm0j>dOEj?(s^qx5)
zng5~X52!=lEmy)T`xVm@hu8`1tpZowqA1mPj0NhK>mJ7<-=WNC(T;xj`+-L6gyQ-+
zc)tF+bc2{&B%a!Jh&D`mCYcOEF=mclCYc1in9nfVEW-=i0^-eZrXiI3f0Oi)gO#}P
zw&eJ;<o-Z%-xi%I@sX5hO^G=)aJa4<`lF+4mH#Q1675?#qD8oiL|({H+1>?zJ5+;f
z!O_^*VJF}V%}>_#RBLg-g@04AkLZnDdq*|8QGTS5+}!lzKrVN)c@K%-sg?px_MJL+
zz{cLP@g%<G<N_}JYGXOWze`+h>+c<QGljn;p+*nSqXvkox9-a)!1`a6(t;L^EU3|l
z0N4EOHX$g;m{wZc6ck`jN3DrG<hXe-24fwi{Obq}dSSG=0AA>L8uvg=V9g;mF&@;_
zhmZw%P|(!pg3oM4t!A0qOI(KYi5M-&Q&-tsMKmb>50@yHEe~_ZY01Sr*kk$M>lN)D
z6u@0M_hu0wacn1?Zn43oW4W-};wE_yXX_ueY+>w$_m5WL-^1ObuP)iu?~n=9_o=*0
z<quRgsF0FG>T^^+B$DRi3FUM&jbmx<+=LQIbCX2XkBzPbIj4EePGZkjTJ~h$rXE!8
z(z?1#WEpK&AdB--jlQcbi5Ybhc68U{O*qzFT<{4=tR}C5M$cs#r7VSN4>kA@mb%Xz
z_Tn@<o`?~NE_`!m?SiJ(YLki@t(}i*+PE^W)Owk;IHSZ*FVNMB_pz6OGmq~o^s?E_
z2;tcLtddBZ6XO%3;ct*EDD83c|0EOD=tp`wG4n437kgc}Nx#}_#mtuoe%@DukLrKz
zdr`0$r?p9PwS2%`JK)8WCb)X?5!|D{e)2~gPQlQ~*O|`|Tp4MTII>-p%8I@@(#m26
z_Jn_l%V0BnP~vGHmFM)z$T5b{xu{PPDW&oY{U^~+S)qj97r}b0jX2+lH4uxusb*XZ
z!KpWPRnpGgM7%lb=k%C76ue2JQt(bU`l@K3UG-61OC&8!pHtM+TAH6ZtHj3Da)ZsB
zjVrTJ)d`oTTP$TXm&&ghJw5#vW2vTq%0C$Wh4>1V$Wsusc{QrcLRG>irNy?n$ii$i
zJ`Y0)g~>4b+liMjF3?BS^GsDc9UY-7Y9o=xk%$&awl<!noJxg`7e-kufwK50ZinX=
zzlIO%?=Jp@!ENyJc{lFTZ=TO#NdKR{)y)Tq(V*`A_BD)U_-LulDbYFlsf3A_mr)0x
zWI0#5PTi@*pm}-0{TNB`^6C@;Cr+GDja%IB#V5Su@ZR!Uf^mQnTzqm5w!xcEs)qzJ
zAnlNYM_Z5dkCK0Tpu6v}c3Cx1?ra|%Z6oaA4l7-P8NyFJGeAG2D*fIVUzRjKF*B#k
z8?Ql{H!4o0m7eB$NS&&*@YHE4GgRK7VpNV%B1T0RPo)};uM`P%`|5R=T`7_lNJxH2
zE6^{mba6|^F+PG*W@qxi8n-SaAKfsQWLN{<k#;0o*LJSwB|A4HPlh*OMnKlwq`X$K
zUYQ);u#{#@Su7%3UNW*)zkV<|y<siOn6p?y#N;7oOBP!nn84TtQ+^iO|H~rhI4oT#
zxYwEYU_^FW*Kv65LP1-0WHE9SrbNqX|91y2SUzJJhma}B_Edb;b8EMOw^+pFWteQr
z5ZG4h96nr|l5824T<_Q&8QupytFHazE<<+aNbXEVt~4a4HXNnK<|vm4&sb<=MRv!w
MjYeVp<JBVO-@7?n&;S4c

diff --git a/app/main.py b/app/main.py
index 215e881..19ed9ad 100644
--- a/app/main.py
+++ b/app/main.py
@@ -414,14 +414,14 @@ async def get_submission(
     pagescales: Annotated[
         str, Form()
     ],  # Scales of Pages  # Annotated[List[Dict[str, float]], Form()],
-    censor: Annotated[str, Form()],
+    ocr: Annotated[str, Form()],
 ):
     """handles submission"""
     print(
-        f"lva: {lva}, prof: {prof}, fname {fname}, stype: {stype}, subcat: {subcat}, sem: {sem}, ex_date: {ex_date}, rects: {rects}, pagescales: {pagescales}, censor: {censor}"
+        f"lva: {lva}, prof: {prof}, fname {fname}, stype: {stype}, subcat: {subcat}, sem: {sem}, ex_date: {ex_date}, rects: {rects}, pagescales: {pagescales}, ocr: {ocr}"
     )
     info(
-        f"lva: {lva}, prof: {prof}, fname {fname}, stype: {stype}, subcat: {subcat}, sem: {sem}, ex_date: {ex_date}, rects: {rects}, pagescales: {pagescales}, censor: {censor}"
+        f"lva: {lva}, prof: {prof}, fname {fname}, stype: {stype}, subcat: {subcat}, sem: {sem}, ex_date: {ex_date}, rects: {rects}, pagescales: {pagescales}, ocr: {ocr}"
     )
     rects_p = json.loads(rects)
     scales_p = json.loads(pagescales)
@@ -454,15 +454,24 @@ async def get_submission(
     censor_status_datas[fileId] = {}
     if fileId not in censor_status_update_events:
         censor_status_update_events[fileId] = asyncio.Event()
-    await asyncio.to_thread(
-        censor_pdf,
-        filepath,
-        dest,
-        rects_p,
-        scales_p,
-        False if censor == "False" else True,
-        fileId,
-    )
+    if ocr == "True":
+        await asyncio.to_thread(
+            censor_pdf_ocr,
+            filepath,
+            dest,
+            rects_p,
+            scales_p,
+            fileId,
+        )
+    else:
+        await asyncio.to_thread(
+            censor_pdf,
+            filepath,
+            dest,
+            rects_p,
+            scales_p,
+            fileId,
+        )
 
     # return {"done": "ok"}
     # print(dest)
@@ -505,7 +514,6 @@ def censor_pdf(
     destpath: str,
     rects: List[List[List[float]]],
     scales: List[Dict[str, float]],
-    secure: bool,
     file_id: str,
 ):
     """Censors pdf and saves the file to the given Destpath.
@@ -549,6 +557,65 @@ def censor_pdf(
     censor_status_update_events[file_id].set()
 
 
+def censor_pdf_ocr(
+    path: str,
+    destpath: str,
+    rects: List[List[List[float]]],
+    scales: List[Dict[str, float]],
+    file_id: str,
+):
+    """Censors pdf and runs OCR
+    The file is converted to Pixels and then recreated.
+    Saves the file to the given Destpath.
+    Args:
+        path: path to the pdf document
+        destpath: Path where the result is supposed to be saved to
+        rects: Coordinates of rectangles to be placed on the pdf document
+        scales: Scales of the rects coordinates for the pdf document
+        secure: weather or not the pdf document is supposed to be converted into an Image (and back) to make shure, the censoring is irreversible
+    Returns:
+        None
+    """
+    doc = pymupdf.open(path)
+    output = pymupdf.open()
+    page = doc[0]
+    npage = doc.page_count
+    for i in range(npage):
+        page = doc[i]
+        if i < len(rects) and rects[i] != []:
+            print(i)
+            wfac = page.rect.width / scales[i]["width"]
+            hfac = page.rect.height / scales[i]["height"]
+            for rect in rects[i]:
+                prect = pymupdf.Rect(
+                    rect[0] * wfac,
+                    rect[1] * hfac,
+                    (rect[0] + rect[2]) * wfac,
+                    (rect[1] + rect[3]) * hfac,
+                )
+                page.draw_rect(
+                    prect,
+                    color=(0, 0, 0),
+                    fill=(0, 0, 0),
+                )
+        censor_status_datas[file_id]["page"] = i + 1
+        censor_status_datas[file_id]["pages"] = npage
+        censor_status_datas[file_id]["done"] = False
+        censor_status_update_events[file_id].set()
+        # THis Costs us dearly
+        bitmap = page.get_pixmap(dpi=400)
+        pdf_bytes = bitmap.pdfocr_tobytes(
+            language="deu",
+            tessdata="/usr/share/tessdata/",  # tesseract needs to be installed; this is the path to thetesseract files
+        )
+        output.insert_pdf(pymupdf.Document(stream=pdf_bytes))
+        # End of the costly part
+        print(f"Page {i + 1}/{npage}: CENSORING DONE")
+    output.save(destpath)
+    censor_status_datas[file_id]["done"] = True
+    censor_status_update_events[file_id].set()
+
+
 def test_function(i: int) -> bytes:
     return b"\x00\x66\x99"
 
diff --git a/index.html b/index.html
index 2051433..e2724f3 100644
--- a/index.html
+++ b/index.html
@@ -134,11 +134,10 @@
             </div>
             <input
               type="checkbox"
-              name="censor"
+              name="ocr"
               id="sec_censor"
               value="True"
-              checked
-            /><label for="sec_censor">Zensieren</label><br /><br />
+            /><label for="sec_censor">OCR</label><br /><br />
             <button type="submit" id="send">Senden</button>
           </form>
         </div>
diff --git a/static/app.js b/static/app.js
index ceea83f..f654998 100644
--- a/static/app.js
+++ b/static/app.js
@@ -276,8 +276,8 @@ function submitPdf(eve) {
   formdata.append("fileId", doc.fID);
   //formdata.append("filename", doc.filename);
   formdata.append("ftype", doc.filetype);
-  if (!formdata.has("censor")) {
-    formdata.append("censor", "False");
+  if (!formdata.has("ocr")) {
+    formdata.append("ocr", "False");
   }
   console.log(formdata);
   submitForm(formdata);