Home

(TITLE OF THE THESIS)* - UWSpace

1. e eeeeee eene nnne 37 SIL Be TranstottidttOb oor taedet e cep odas ede nn eicaa Desc da el debes Mou COS ied ial 39 S12 Ea DranstoOLtidttoDteiuedoioncotec qeudeeaeece itn dutundotee tuvo udoc c itu dutem dot 40 2 2 stretching a Letter in Persian Nastalqr ui oio to eR rene e Ex ERE M ERE 42 Chapter o The Nastalo Font sic e ede eoa ose dpa dede Sud a bomi edd 50 6 1 Why Develop a PostScript Type 3 Font eeeeeessseeeeeeneeeenn 50 6 2 The Pont Used inthis ResearclIis ou oe eode tet Serra aae ee N peg d aues 50 G S COnVersiODm DOOlS sceccconeeta b diei neces deis S 51 Vil Od Cony enime tla tO T VDO uttter rome A ner robados issuibese cit tec 5l 6 Glyphb TdeUttfIeablOfL unio edite sn taut sete eror e peg odd dte taut 52 0 6 Nami The CIT BS ec i eret DDR Ree Rude cu m pda Rae Fc Edb m pisa LAS 55 6 7 Calculating Font MGIIiCs oos em cov nea Sava A E uad Ex 59 6 8 Identification of Stretchable Curves c cccccccccccccsseessssnsssscssesssseceeccceeseaseeeees 60 6 9 Impleimentule The Font donec es tene CREER dura etas co arctan cm o idum bu omit 61 ERCE AONE dS Scouts Mitos erstes et UE 66 6 10 The Problem of Misplaced Curves eeeeessseeeeeenneerennnn 73 Chapter 7 Conclusions and Future Work cccccssscssscccceecceeeeseeeceeeeeeeeeeeeees TI FNIT E TI deel TASC OF C OntriDUttODS ausestu m docete se eto RM Motu as E ef eR IE dun 78 7 2 Lessons E edrnetdus susto
2. Jim Che c He cz Khe 3 Ghaf and J Ne This rotation occurs to draw the curved shape of the letter Figure 22 shows letter He As shown in the figure the nib head rotates up to 20 degrees 30 Figure 21 Nib Head Using the Power Style and Solid Mode in Letter Te f Figure 22 Nib Head Using Power Style and Rotate Mode in Letter z He Stretching a letter is done using the power style and in the solid mode and as for the Arabic Naskh style the angle of the nib head is about 70 degrees from the horizontal axis It is 31 also necessary to mention that in Nastaliq the part of the letters that are written using rotate mode never get stretched however this might happen in Arabic Naskh 32 Chapter 4 Research Leading to this Thesis 4 1 Original Goal Our main purpose was to modify ffortid in order that 1t could support Nastaliq writing obeying calligraphic rules with dynamic stretching respecting both vertical and horizontal movements of letters 4 2 Method to Achieve Goal For this purpose I decided first to write a user s manual for a new version of ffortid called NAS that would handle the Nastaliq writing This user s manual would serve as a requirements specification 4 3 The Process of Preparing the User s Manual for NAS It was first a little bit unusual for me to start with just writing the user s manual but after many discussions with my supervisor who played the custo
3. X10 y h2 _ 13 Y12 The line from M12 to Mi3is X X12 X13 X12 M10 M11 M2 M43 Y11 Y10 es py gt gt x Xio Yo x X5 ty12 11 410 13 412 Y11 Y10 Y13 Y12 X X t 10 y 11 X10 12 y 13 X12 Y11 Y10 Y13 Y12 X11 X10 X13 X12 Y11 Y10 Y 13 Y 12 X410 X Yjo Y p E Y11 Y10 Y13 Y12 10 10 X11 X10 X13 X12 Y39 Y12 gt Xr J A 1 Ay with A being the parallel to Mjo M11 passing through the point Myo 45 A being the parallel to M45 M45 passing through the point M 3 y ho 1 Y10 X X49 h X41 X10 y hsgtv fg r1 X X13 X13 X12 CIT or Y13 Y 12 X49 h Y49 X13 Y13 v gt y uo 10 h Yi0 yy 413 tY13 J Y11 Y10 Y13 Y12 X11 X10 X13 X12 Y11 Y10 Y13 Y 12 Yi Yig Xv xig 10 th M0 y 2 ay X13 TY13 V amp y X41 X H Y11 Y10 niri X10 h T Yo Seu X11 X10 X13 X12 Step 5 Calculating Metrics for the stretched version of shifted curve Note again that X7 p g g means the x value of and Y means the y value of I Now we shift the result curve with the space equal to and angle a and replace Xr by X 1 to get the metrics of the shifted curve with U X1 X11t1 cos a X12 Xizd cos a Y11 Yudl sin a Y11 Yu sin a X12 Xi2tl cos a X13 X13 l cos a Yi2 Yiztl sin a Y13 Yi341 sinta The resulting X will be X Y11 l sin a Y 19 l si
4. 0 cos a 1 Cz X j C2X13 l cos a 7 Xiz l cos a With the same approach Y 22 Y22 l cos a The proof shows that for every B zier curve if you stretch it and then shift it with U I cos a l sin a the result is the same as shifting with U first and then stretching it using the stretching technique of Abdelouahad Bayar and Khalid Sami This result proves that with this method even after the stretching the stretched curves of a glyph maintain the initial distance between parallel curves according to the size of the nib head its angle a with the horizontal axis and its length It is important to mention that even when adding or decreasing constants to the x or y value of the points in the initial stage of stretching Abdelouahad Bayar and Khalid Sami s method continues to be valid A similar proof shows that the Eg and Eq transformations end in the desired results The stretching for Persian Nastaliq uses almost the same method However for the IranNastaliq font used in this research it was concluded that the vertical shift would not be necessary So we can set v to be zero and work just with A 49 Chapter 6 The Nastaliq Font 6 1 Why Develop a PostScript Type 3 Font Recall that the main purpose of this research was to make it possible to typeset Nastaliq using ditroff ffortid and a dynamic font The main point is that ditroff ffortid uses fonts in PostScript Types 3 and 1 But only Type 3 fonts are
5. C3 M22 1 C2 J C2M23 6 9 1 Exceptional Cases Be Pe amp Te and amp Se and not stand alone version of Ne amp Ye and is Ee When stretching such a letter the dot or Hamza that is shown in letter c and makes it to look different from leter s should be shifted by half of the total stretch amount to the right This shift should always place the dot or Hamza at the horizontal middle of the stretched part of the glyph 66 4 135 540 2 135 270 Figure 38 Stretching and Component Shifts Letter Be ii Letters lt Kaf lt Gaf 67 CB All Kaf7 All P R ra Figure 39 Stretching and Component Shifts Letter Kaf As shown in Figure 39 there are two closed paths in this glyph which are connected to each other at point R The closed path located at the left side of R is the main part that needs to be stretched and the right path is just the identifier shape for this letter which identifies if this letter is Kaf or S Gaf In this case the identifier just needs to be shifted by the amount of the stretched letter For the letter S Gaf as shown in Figure 40 the identifier shape contains two parts which means both of them should be shifted 68 P4 p it A I nN a U 7 Figure 40 Stretching and Component Shifts Letter gaf The stand alone version of letter has a Hamza that needs to get shifted by the stretch amount It is imp
6. Mt 2 XizT l cos a Yi2tl sin a M13 X134 l cos a Yia7 sin a and M1 M10 h 0 Mis Mis 0 v Mi1 1 C1 Mio Cy Miz 1 C2 Co Miz 47 Step 6 Proving that the stretch shifted version 1s the same as shifted stretched version Now we need to check if the stretched curves have the same distance from each other In this case we need to check if the control points of the stretched curves also respect the U transformation or not M19 Mto h 0 Mt 3 M 13 0 v Mt r Z 1 Ct1 Mtio Ct Mt z 1 C 2 J Ct2 M zs gt Mt o X10 l cos a Yio l sin a h 0 M3 X134 l cos a Yi3 l sin a 0 v Mt r 1 Ct M io Ct J M z 1 C2 J Ct2 Mtis Xt10 X10 1 cos a h X10 l cos a h Yti10 Y10 1 sin a Yio l sin a X 13 X13 1 cos a X13 1 cos a Y 5 2 Y13 l cos a v Y13 l sin a v Ugy X 1 C1 Xo CtiXE and remember X11 X10 G Ix 410 X12 l C2 UNE X11 l cos a X190 1 cos a I l cos a X40 1l cos a So C5 E 48 X12 l cos a Iy l cos a _ and Ct X13 H cos a lx l cos a C So Xt r 1 Ct Xt1o Ct1 Xp 1 Ci X10 l cos a h C X 0 cos a 1 C1 Xig h Ci Xjy 1 cos a Xir l cos a With the same approach Yt1r Yi l sin a At last Mt1 z 1 Ct2 J Ct M43 So Xt z 1 Ct2 Xt Ct X3 1 C2 X L cos a C2 Xis
7. exists such letters in these 2 sections choose them using the golden rule and stretch them by the equal amount up to 11 points so that the line 1s filled If these two sections do not both contain the suitable letters check section S2 S5 82 S4 S4 S1 respectively If the stretching in none of the group of sections given above would not fill the line it will mean than we need more than 2 stretches Check sections S1 S3 S5 If each contains a stretchable letter by the golden rule choose a stretchable letter in each section and stretch them by the equal amount up to 11 point as much as the line is filled If the line cannot be filled and there is still spare space there is nothing more we can do So we move to the next line 4 2 2 Manual calculating In manual calculating the user will have the choice to ask for a maximum stretching mode The applied rules and uncommon cases and have been introduced in chapter 3 already Notice that manual stretching can still be applied to stretchable letter defined in 4 2 1 3 5 References l Learning Persian Calligraphy Fazaeli Habibbollah Publisher Yasavoli 102 Rules of Calligraphy by Amirkhani Adaabol khatte Amikhani Amirkhani Gholam Hosein ISBN 964 6271 05 7 Publisher Organization of Persian Calligraphers 7 Edition A look into cheminstry in Nastaliq Falsafi Amir ISBN 964 306 092 6 Publisher Yasavoli 15 Edition Learning Nastaliq se
8. letter E faces such a condition but since it is mostly contained in section 4 than section 3 we assume it 1s in section 4 not 3 it can be considered as a member of the section which contains a larger vertical portion of the letter If the letter 1s equally in two sections it can be considered to be in any of those two LJ LE we i id i When we just want to have one stretched letter in a line the best place to choose its location is section 3 If the letters in this section are not stretchable our next candidate is 98 section 4 and again if section 4 does not contain such a letter the next candidates are section 2 5 and 1 in that order If we want to choose 2 stretched letters 1n one line the preferred situations are 3 5 If it is not possible to choose stretched letters from those sections the next candidates are 2 5 2 4 and 4 1 in that order Rule 11 When fully stretched for some letters such as the beginning of the letter should be replaced one dot higher than usual 3 Chapter 3 Basic Use Cases 3 1 An overview of stretching In this section the instruction for stretching letters is provided This software provides two different Ways to stretch letters Automatic and Manual 3 1 1 Automatic To obtain an automatic stretching the user has to choose the both side justified ec paragraph by the ad p command In this way the software automatically choose
9. with all stretching amounts of all shapes of all glyphs rather than also making the changes to ditroff ffortid to handle this font As a result the user s manual for NasTroff was left incomplete However the developed font was designed to meet the requirements of being used in NasTroff Since the font is written in PostScript it can be used even without any word processor 1 4 Notational Conventions In order to make the content of this thesis easier to understand here I provide the list of notational conventions Different fonts have been used to show different concepts e Times New Roman is used for the body of the article e Bold Arial is used for chapters and sections names e Cambria Italic is used for mathematical formulae and variables e Palatino Linotype is used for showing rules in Persian Nastaliq 6 e Comic Sans is used for the formal grammar of the regular languages e Calibri is used for Algorithms e Each Persian Letter is first shown in Persian followed by its English name in parenthesis e For two points Kz K2 the notation K lt Kz means Ki is located at the left side of K2 K1 gt K2 means Ki is located at the right side of K2 and Ki K2 means Ki is located in the same horizontal location as Kz Also Kz gt Kz means that K is either located on the right side of K2 or in the same horizontal location as K2 and K lt Kz means that Kz either located on the left side of Kz or in the same horizontal loc
10. B h v gt Ej B h v Note that represents the horizontal stretching amount and let v represents the vertical stretching amount The transformation Ej stretches curves in p1 The details of its definition are given below Let By Mj9 M11 Mj5 M35 and By Mo9 M21 M22 M33 be two curves in f4 Let h v e 0 h x 0 vnl 5 1 1 Epe Transformation B Ej B4 h v if and only if the control points of B are M29 Mio h 0 M73 Mis 0 v Mj 1 c1 Mz0 cJ M5 1 c2 C2Mp3 with c4 c satisfying My 1 c1 M10 cJ M 2 1 c2 I c2M33 39 where j Mj9Mq1 N Mi2Mi3 and J A N A with A being the parallel to M M passing through the point Mj and A being the parallel to M45 M45 passing through the point M33 An example of a stretching using the function Ej is presented in Figure 26 Figure 26 Stretching a Curve Belonging to the Set 6 with Ej B4 B 5 1 2 Eg Transformation Let Eas be the stretching function defined as follows Esp B2 X 0 hm x 0 Un gt B2 B h v gt Eg B h v The transformation Ear stretches curves in 62 The details of its definition are given below Let By M1o M11 M45 M35 and B M59 M21 M22 M55 be two curves in 5 Let h v 0 hm x 0 v 40 B Eas B1 h v if and only if the control points of B are Mz Myo h v M23 Mia Mz 1 c1 Mz0 cJ Mzz 1 c
11. Base line In Nastaliq the baseline of each character is determined according to the position of the character in the word containing it and whether it is going to be connected or not As a necessity it is the case that we start every maximal group of connected 92 letters from a height higher than the original baseline and we come down while writing on a slanted line so that the last letter in this group is either tangent to the baseline or under it due to the nature of the letter 99 Shape Changes In Nastaliq Letters o and might change shapes to a shape different from the four ones already introduced When applying Keshidah most letters will become just a little bit more stretched but some letters such as letter o lose all their teeth This case happens only to letters UH and c The Iran Nastaliq font used in this software does not support the stretch glyph however by a modification to the PostScript code of existing letters it 1s possible to produce the desired shaped Ptrn A pre processor which would do letter form and ligature identification on letter only input to yield output with each glyph to be printed Also shape changes which may occure are handles in this sections ditroff troff s processor which handles the left to right languages ffortid troff s post processor which handles the right to left languages including stretching psdit converts the ditroff or
12. Has a Vertical Movement From Its Previous One on The Baseline a Figure 4 A Sample of Two Words in Persian Written in Book Typing Typeface on the Left Side and Nastaliq on the Right Side Another issue that makes Nastaliq different from many typefaces but still similar to others is stretching The same as Arabic Naskh and Hebrew some letters can be stretched in Persian Nastaliq Figure 5 shows stretching in different languages such as Hebrew Arabic Naskh and Nastaliq This stretching can be used for several reasons the most important one is to replace the extra space at the end of lines to achieve line adjustment which is a common typography technique In English this justification can be achieved by increasing the distance between the words in order to align the along the margins This method does not work for Nastaliq since in Nastaliq words should have a constant distance from each other As stretching could be implemented respecting several calligraphic rules the main stage of this research was to prepare a set of rules for stretching in Persian Nastaliq In addition while Nastalig has up to four shapes for each letter the shape of these shapes can change as a function of the letters before and after them Therefore a simple Persian Nastaliq typeface might contain hundreds of glyphs in order to be able to handle all possible letter combinations 2 3 B C Figure 5 Different Examples of Stre
13. I had to rename the 878 glyphs The glyphs in the original fonts had random names which make it hard to use them normally to write PostScript programs that would print Persian text A sample of naming in the original file has been shown in Figure 31 As can be seen the name provides no information about the glyph Therefore I had to identify 52 each glyph and its possible context The name of any glyph is an encoding of its letter its position and all of the possible contexts I defined a regular expression for naming the glyphs by this encoding It was easy to envision either a person or a software application choosing any glyph by determining its name given its letters connectivity and current context CF Ashaf_Jim Che He Khe IY a r pe A Figure 28 Connect Following Letter Ghaf Followed by One of the Letters z Jim G Che e He c Khe CF Ghaf Mim None VA Figure 29 Connect Following Letter 3 Ghaf Followed by Letter Mim Which is Not Followed by Anything 53 CF _Ghat_Mim gt All VY JN Figure 30 Connect Following Letter Ghaf Followed by Letter e Mim Which is Followed by Another Letter Figure 31 Different Combinations of Letter Ghaf 54 afi157449 Figure 32 Random Naming for Glyphs in the Original Font Letter Ye The possible connectivities are SA for stand alone CB for connecting both CP for connecting previous and CF for connecting following 6 6 N
14. We assume the user uses the free editor program called Textedit which is pre installed on all Mac OS X systems and will build the result using Troff the plan 9 version modified by the author of this article NAS was decided to handle these as follows 1 A Persian font would provide the different forms of each letter as independent characters and each character that is to be connected on any side would be designed to be flush to the bounding box on that side at precisely the same place relative to the baseline 2 A pre processor called ptrn would do letter form and ligature identification on letter only input to yield output with each glyph to be printed be it a form of a letter or a form of a ligature The letter only input would be according to a standard encoding for the language being processed and the output would be according to the font s encoding for the glyphs Thus ditroff would format input consisting of the glyphs to be printed If the input to the pre processor has diacritical marks then they will be translated into their glyph codes surrounded by instructions to place them in the proper vertical position with respect to the character with which it 1s associated 87 3 The ffortid postprocessor would be modified to stretch connections to last letters of words and or lines 1n order to achieve one kind of Keshidah Figure 1 describes the flow of how this software works The user inputs Unicode text in the edit
15. able to handle dynamic stretching This difference occurs because the Type 3 PostScript code of a glyph can contain parameters while for type one there can only be constants Dynamic stretching means that only a single parameterized outline per stretchable character is needed the actual outline of any stretching of a letter can be determined at printing time by the PostScript interpreter A Type font has a fixed outline for each character A separate outline is needed in the font for each stretching amount of each stretchable form of each letter Therefore I developed a PostScript Type 3 font however I did not make it dynamic but containing all the pre compiled variations of all stretched glyphs To build this font I had to find an existing Nastaliq font and convert it to Type 3 6 2 The Font Used in this Research The Nastaliq font used in this research is called IranNastaliq It is a TrueType font and is free to download and use in Microsoft Word under Windows However when using this font in other applications or operating systems several problems exist For instance when using Microsoft Office with Mac OS X only the stand alone version of each letter shows up and letters cannot connect Several fonts were reviewed in order to chose the best 50 to be converted to a PostScript font These fonts include the IranNastaliq WM_Nastaliq by Maryam 8 and NamehNegar 9 The reason that IranNastaliq was chosen is that this font uses abs
16. in Section 2 2 1 there are two kinds of stretching in Persian We can either stretch the connection between two letters or a letter itself respecting the rules introduced in Section Chapter 3 Note that the amount of stretching 1s a multiple of the width of a dot Figure 18 shows a stand alone version of letter 2 Be with a size of 5 dots A dot is shown at the bottom of the letter in this figure Note that for the font used in this research the initial size of the unstretched version of many letters such as Be Pe Te and Se is a little more than 5 dots In order to calculate the exact size of a dot used in this font I divided the total length of such a letter by 5 and calculated the exact amount of the dots used in this font which is 135 points in PostScript Therefore a letter s length might not look exactly equal to the size we 21 expect Figure 19 also shows a stand alone version of letter Be that is stretched by the amount of four dots SA Be 6 Figure 18 Unstretched Stand Alone Letter Be with Size 5 Dots The Typical Dot Can Be Seen in the Bottom of the Letter 28 SA Be9 d Figure 19 Stretched letter Be of Figure 18 by the Amount of 4 Dots Use of the nib head in Persian Nastaliq is not the same as in Arabic Naskh There are two different styles of writing with the nib First is called the weak style which is seen mostly in the beginnings and the middles of some letters and
17. is aligned If there still exists spare space in the line If there are no stretchable letters in section S3 check sections S4 S2 S5 and 1 in the given order If the line still has spare amount we would undo any changes done in section S3 NAS first checks if it is possible to make the changes if so it makes the changes if not it will move forward to the next step without any changing in order to save time There is a golden rule provided here to make our job easier to find the suitable letter for stretching in a section containing more than one stretchable letter This golden rule is just a pseudo code for implementing rule 6 in chapter 2 4 The golden rule says if there are more than one stretchable letters in a section and we need to choose one to stretch we would stretch the first stretchable one The only case this does not apply is if on the above line in the same paragraph there exists a stretched letter exactly located at the top of 101 our choice If this is the case we move forward the next choice in the section If the next choice also is problematic or there is no other choice in the following sections we will move forward with the choice we made first Now we know that for the line we are in more than one stretch 1s needed So by rule 10 in 2 3 check sections S3 S5 find the stretchable letters respecting 4 2 1 3 and 2 8 rule 8 in these two sections each should contain at least 1 stretchable letter if there
18. is either X or X2 then we need to check where the curve ends a d X1 Ps m l Az Figure 36 A 4 Point Stretchable Curve Glyph Let Cp be the current control point If Cp X1 or Cp lt Xz then do nothing don t move the curve or line starting Cp Else If Cp X1 or Cp X2 then Call the end point of the curve starting at Cp Ep If Ep X1 or lt X2 then do nothing don t move the curve or line starting Cp If Ep X3 or Ep X4 then apply stretch 1 7 Else If Cp gt X3 or Cp gt X4 then add the stretch amount to the x value of start and end point of the curve move the curve s or line s starting at Cp by stretch a ae P9 1 amount horizontally 8 Else If Cp X3 or Cp Xasthen 9 Let Ep be the end point of the curve starting at Cp 10 If Ep gt X3 or zX4 then add the stretch amount to the x value of start and end point of the curve or line move the curve or line starting Cp by stretch amount horizontally 62 11 IfEpzXsor Ep X4 then stretch 2 12 Else if Cp Xs or X4 add the stretch amount to the x value of the start and end point of the curves move the curve or line starting Cp by stretch amount Method Stretch 1 Add the stretch amount just to the end point of the curve the start point does not get shifted but the end point gets shifted by the stretch amount horizontally Method Stretch 2 Add the stretch amount just to the start point of the curve the start point gets shift
19. letter can be stretched by an amount of from one to five dots Rule 6 A letter stretched by an amount of two to five dots should not appear at the beginning or end of a sentence Rule 7 No two stretchings in two neighboring lines should be located vertically close to each other If there are no other choices it is permitted to disobey this rule Rule 8 No stretches may appear at the beginning and end of a line However if there is no other choice and we really need to stretch we can stretch in these situations Rule 9 24 Stretching is acceptable in the following eight conditions and not in other contexts a b gt 4 after 7 He or 7 He as in ou and after sad Or Ta as in e 4 and a a cl c d e f g h Rule 10 J after uis A Ein as in ra and adi a A Z CA after Fe or Ghe or Fe or Ghe as in and after Mim as in 4 after gt Ha as in s after 9 Ha as in Mesh and and E 4 r S J after Ha as in uU and The letters o and c cannot be stretched in the following conditions 25 a before t Jim Che He or Khe P i Che He or Khe or Fa Che He or Khe Re b before another stretched version of the same letter c before C stom TA or A iim d before a ai e before l Ye Rule 11 In every line it is suggested by Nastaliq Masters to have 1 stretched letter or two halt stretched ones instead of
20. need to find out which letters are stretchable According to the font we use in NAS most of the letters are 9 stretchable Letters such as o aLi are stretchable if it is either in the stand alone the connect before or connect both position Letters such as 100 esie Ke g hho ate stretchable only in connect both and connect after positions These letters are stretchable 1f they obey rule 2 4 8 4 Since we prefer to have not many stretched letters and according to rules 10 in 2 3 we would rather to have stretched letters in specific places in a line we also need an algorithm to find the suitable letters and stretch them i If the existing line is the last line in a paragraph 1 e it contains a period and there is nothing after the period automatic stretching would not apply to this line if not move to next step Divide the length of line into 5 section call them S1 S2 S3 S4 S5 from left to right which means that the most left section is S1 and the most right one is S5 If the majority of a letter is in one section but some of it is also in another section We would consider the section containing the majority as the main section containing the letter If it 1s the case that a letter 1s equally in two sections we would consider it to be in the first one Check Section S3 if there exists a stretchable letter which would not contradict rule 8 in 2 3 stretch it to the maximum amount up to 11 points so that the line
21. outline of any character is a series of elements each of which is a line a cubic B zier curve a circle or an arc Daniel Berry stretches the B zier curves which do not affect the nature of the letter Figure 6 illustrates the stretching of a cubic B zier curve Figure 6 Stretching a Four Point Curve Using the Method of Daniel Berry 2 To stretch a 4 point B zier curve by A units simply add A to the x values of the two right hand points and to all points to the right of the left hand one of these It is a problem to stretch through the shared end point of two B zier curves whose tangents are the same at the shared point As shown in Figure 7 adding A to the x values to the three rightmost points of the right hand curve in order to stretch between the first and second point introduces a corner into what was a smooth meeting at the shared point 12 Shared end point A AN wont BN E puc s eo Cornered shared end point Figure 7 Cornered Shared End Point Caused by Stretching As shown in Figure 8 the corner can be avoided by preserving the slope in the left hand tangent of the right hand curve by increasing both the x and y values by amounts consistent with the slope of the tangent However now the right end of the combined curve does not have the same y value as before A proper solution requires redesign of the two adjacent B zier curves into one or three adjacent curves so that the place
22. text to achieve right to left text in ditroff ffortid is as given below To understand this algorithm better Daniel Berry and Johny Srouji first explain how ffortid works The main job of ffortid is to reorganize the characters so that the text is in visual order in which the text of each language is written in its own direction and the flow of the single directional chunks in each line is consistent with the current document direction The 8 current document direction 1s left to right or right to left as the user decides To be more specific current document direction is usually the direction of the main language used in the document For example typically an English book 1s a left to right document even if it contains a lot of text in right to left text ffortid works by totally reformatting each line as a function of the current document direction and the direction of each character At the top level of abstraction it reads the characters of each line delimited by the end of line marker and permutes characters so that the uni directional chunks of a line flow in the current document direction while the characters in each chunk flow in the chunk s own internal direction for each line in the file do if the Current Document Direction is left to right then reverse each contiguous sequence of right to left characters in the line else the current document direction is right to left reverse the whole line reverse each contiguous sequenc
23. the same time For future work it 1s recommended to work on a program that can handle both vertical and horizontal movements in TrueType fonts Another important issue is handling the problem misplaced start and end points of curves in the glyph as mentioned in Section 0 The solution for this problem is changing the stretchable B zier curves so that the start and end position of each curve is in a suitable place for stretching This change requires cutting a B zier curve into two new curves given a break point or combining two adjacent B zier curves The combination is usually helpful to reduce the number of curves in the layout of the glyph in order to achieve a better look Figure 50 illustrate two proper curves B2 and D2 As shown in this figure by breaking the curves B and D in Figure 47 we can achieve two curves 52 and D2 which their start points and end points have much less vertical distance 79 SA Kaf Figure 50 Breaking Curves B and D in Figure 47 to Achieve Proper Curves in letter lt Kaf 7 4 Breaking a B zier curve into Two B zier Curves Given its four Points and a Point Placed on the Curve It would be enough to show that if any B zier curve is broken at any point into two curves these two curves would be B zier curves It is clear that if you break any polynomial curve the two curves are still polynomials with the same degree We know that every B zier curve is a polynomial of degree 3 Since every polynomia
24. 2 c 2M5 with c4 c satisfying My 1 c Mjo cJ Mi 1 c2 I c2M33 where 1 M1oM11 N Mj2Mi3 and J A N A with A4 being the parallel to M419 M41 passing through the point Mj and A being the parallel to M45 M45 passing through the point M33 An example of E stretching is given in Figure 27 Figure 27 Stretching a Curve Belonging to the Set f2 with E B Bz 4 Abdelouahad Bayar and Khalid Sami divide the total stretch value of a letters into two parts and pass each half to one of the two functions E and Epe 5 2 Stretching a Letter in Persian Nastaliq Abdelouahad Bayar and Khalid Sami talk about how to stretch a B zier curve with its control points and an amount of stretch so that we respect the constant size of the nib head For a quick review Assume that B is a B zier curve with four control points M1o M11 M1z M13 They provide a formula which gives us 2 which is a curve respecting the constant size of the nib head Given four initial control point M o M M 2 and Mj3 we can get the stretched curve s control points M29 Myo h 0 M73 Mis 0 v Mj 1 4 Mp9 cJ M2 1 c2 c2M33 with c1 c satisfying My 1 0 Mio cJ Mi 1 c2 I c Mas where 1 M1oM11 N M12 M13 and J A N with A being the parallel to Mio M11 passing through the point Mj and A being the parallel to 42 M45 M43 passing through the po
25. Cp Xs or Cp Xe then move the curve or line starting Cp by the stretch amount horizontally Else if Cp Xs or Cp Xsthen Call the end point of the curve starting at Cp Ep If Ep X3 or Ep Xzthen stretch 5 Else Stretch 6 Now let us see how these stretch methods are different From Sections 5 1 and 5 2 we remember Bayar Sami s transformation 64 Given 4 initial control point M o Miz M12 and M13 we can get the new stretched curve Control Points For stretch 1 Apply these changes to Bayar Sami Formula Mz0 M10 M23 Mis StretchAmount 0 M21 1 C1 M20 Cy M22 1 C2 J C2Mz3 For stretch 2 Apply these changes to Bayar Sami Formula M20 M10 StretchAmount 0 M23 Mis M21 1 C1 Mzet Cr M22 1 C2 J C2M23 For stretch 3 Apply these changes to Bayar Sami Formula M20 M10 M23 Mis StretchAmount Z 0 M21 1 C1 M20 C1J M22 1 C2 J C2M23 For stretch 4 Apply these changes to Bayar Sami Formula M20 Mio StretchAmount 0 M23 M13 StretchAmount Z 0 M21 1 C1 Mz0t Cy StretchAmount Z 0 65 M22 1 C2 H C2Mzs StretchAmount Z 0 For stretch 5 Apply these changes to Bayar Sami Formula Mzo M10 StretchAmount Z 0 M23 M13 StretchAmount 0 M21 1 C1 Mz0t Cy StretchAmount Z 0 M22 1 C2 J C2M23 StretchAmount Z 0 For stretch 6 Apply these changes to Bayar Sami Formula M20 M10 StretchAmount Z 0 M23 M13 M21 1 C1 Mzet
26. MT END 74 Figure 47 Letter lt Kaf a Glyph with Misplaced Start Points of Curves Shown in Earr Wires ERN NR M E eieat sented anna dewatered so tad ete eset raat eed st 75 Figure 48 Stretched Stand Alone Version of Letter lt Kaf introduced in Figure 41 Wate C Uf Ves A EO C SERE EC NCC tuere terga au etam ebore Dota dusbt nen tO cen as ea cu Pase UNE 76 Figure 49 Stretched Stand Alone Version of Letter lt Kaf introduced in Figure 41 With eurves b and D StEete Ded ues ise tita Sce oro eoo Ion sepas eate ob ducta ede EU doa tdt 76 Figure 50 Breaking Curves B and D in Figure 47 to Achieve Proper Curves in letter AX cope 80 Table List of All Programs Developed to Make the PostScript Font Containing All Serch GIP IS srun EET OD TITO TT 12 Xl Chapter 1 Introduction 1 1 Persian Writing The spread of computers around the world has resulted in the need for word processing software in many languages One of these languages is Persian which is widely spoken in Iran Afghanistan Tajikistan Uzbekistan and Bahrain and has a status of official language in the first three countries under different names 1 The Persian alphabet has 32 letters and it 1s written from right to left Some of the vowels are shown as letters such as s o u a and c 1 y and some others are not written Each Persian letter can have up to four different shapes according to the position it has i
27. PEDO POPU CD unosi oou a ol a feoda andate bets nade Rc od o 8 PDA ae Be C m 9 P N IG M rero eT LM W Oid c 10 2 4 Persian Only PLO Otani coss peek a tegnss coL Uso tn aan EUREN S watt epar g ue rane 10 PAAR T Ie RT retrain en rear a eee arn meme treet aaa creer ean rate 10 2 2 Diferent V CTSIONS OL Stretching 11 2 3 Pre complied Variations and Dynamic Stretching esses 11 vi 24 Dynamic Font tor stretching letter Serriera reifen 11 2 5 Problem with Daniel Berry s Solution ccccccccsssssseeeeeeeeeeeeaeeseeeeeeeeeeaaas 15 2 6 Abdelouahad Bayar and Khalid Sami s Observations 17 Chapter 3 Rules and Instructions for Persian Calligraphy and Nastaliq 22 3 1 Persian Nastaliq and Different rules for typesetting esssesesss 27 SEM M H eS cei et 22 3 3 How to stretch a letter in Persian Nastaliq essere 2 Chapter 4 Research Leading to this Thesis eeeeeseseeeeeeeee 33 Al CGS Mal CIO ale ius obdoicttut bien hana lores ctt ubbdece saa ph ones ct E dete RA 33 2 2 Method to Achieve GOA xe csi c eee tio euo cca a a 33 4 3 The Process of Preparing the User s Manual for NAS 2 0 ecccceeceeeeeeeeeees 33 44 Tescopme the Researcher eei UE ines Epi Un atti mdasesieauenie aie 35 Chapter Modeling the stret BUDE rss A 36 5 1 Keshideh the Mathematical Model
28. Stretching of the Stand Alone amp Qaf in Daniel Berry s Method In the Arabic Naskh style the nib head behaves as a rectangle of length and width l 6 3 This rectangle moves with a constant inclination angle of about 70 degrees from the baseline A nib s lead with 12mm is shown in Figure 15 Figure 15 Nib Head in Naskh Style Figure 16 shows how a correct stretching should look As can be seen the thickness of the stretched letter is constant and equal to the length of the nib head Notice that this thickness 1s not calculated vertically but along lines that have a 70 degree angle from the baseline 20 Figure 16 A Sample of Correct Stretching in Arabic Naskh Respecting the Constant Thickness of the Nib Head 21 Chapter 3 Rules and Instructions for Persian Calligraphy and Nastaliq 3 1 Persian Nastaliq and Different rules for typesetting In order to make a model for handwritten Nastaliq it is necessary to provide a set of rules that should be respected In this chapter many rules have been gathered which would be useful in order to typeset Persian It 1s important to know that there are also many other rules existing for Nastaliq but the ones provided in this chapter are the only ones important for the purpose of this thesis Another important point to mention is that the master calligraphers of Nastaliq have provided exceptional cases for each rule they have mentioned in order to make the text look better an
29. The Problem of Stretching in Persian Calligraphy and a New Type 3 PostScript Nastaliq Font by Shahab Mohsen A thesis presented to the University of Waterloo in fulfillment of the thesis requirement for the degree of Master of Mathematics in Computer Science Waterloo Ontario Canada 2009 Shahab Mohsen 2009 AUTHOR S DECLARATION I hereby declare that I am the sole author of this thesis This is a true copy of the thesis including any required final revisions as accepted by my examiners I understand that my thesis may be made electronically available to the public 1i Abstract This research is about a typeface for implementing Persian calligraphy called Nastalig The main purpose for developing this font was to handle stretching of letters in order to achieve line justification through a dynamic font Therefore a PostScript Type 3 font was developed However as the research progressed it came clear that Nastaliq s stretching cannot be implemented in a dynamic font Therefore the research s purpose changes to implementing a font containing all the needed glyphs of all needed stretchings of all stretchable letters to allow achieving line justification For this propose a mathematical formulation to model handwritten Nastaliq was necessary The result was a PostScript font containing more than 1200 glyphs To make it possible to use this font in the future a regular expression grammar was developed to identify a
30. after each other in a line we should consider a space of a dot between them If the ending part of the previous letter or the starting part of the next letter 1s thin they can be shifted closer to each other until they look connected Rule 3 Page 92 Learning Persian Calligraphy There are three reasons to stretch letters which the third one is important for us in NAS since the first two needs human interaction to choose what is more beautiful I To look more beautiful II To avoid confusing between similar words III To justify lines and fill empty spaces Rule 4 Learning Persian Calligraphy A full stretch is between 9 to 11 points and a short one 1s from 4 to 5 points Rule 5 Learning Persian Calligraphy 95 A Full stretch does not appear in the beginning or the end of a sentence Rule 6 Learning Persian Calligraphy No two stretches in two neighbor lines should be located vertically close to each other Rule 7 Learning Persian Calligraphy Neither full nor short stretches may appear at the beginning and end of the line Rule 8 Learning Persian Calligraphy Stretching is acceptable in the following 7 conditions Sm S F p gt After as in Aud s uA After i l 2 i After i as in La a3 After as In After A as In 96 bye Ot After as in As z ai Rule 9 Learning Persian Calligraphy We also are not allowed to stretch letters and c in following cond
31. aming the Glyphs To help understand the naming process Figures 33 34 and 35 provide three examples of letters with their complete names Figure 33 shows the connecting previous position of letter Pe stretched by 1 dot to the total amount of 6 dots that is written after one of the letters z Jim e Che z He e Khe Sin Shin o Sad Zad Ta amp Za e Ein Ghein Mim or Ha First comes CP for connecting previous Then comes the previous context 55 Jaim Che HelKhelSinlShinlSadlZadlTalZalEinlGhein IMimIHa then comes the letter Pe Finally comes the stretch amount 6 If the glyph had a following context it would come after the 6 CP JimlChelHelKhelSinlShinlSadlZadl TalZalEinlGheinlMimlHa Pe6 ee e 9 Figure 33 A Sample Naming for Stretched connecting previous Pe Figure 34 shows the shape of the connecting following Be with no stretch that is followed by one of the letters 2 Be 2 Pe Te or amp Se which is itself followed by one of the letters 2 Be Pe Te or 4 Se Figure 35 Shows the shape of the connecting both Te with no stretching amount that is placed after any letter and followed by a e Mim that is not connected following to anything Its name is CB All Te MimQwNone The All here means that the Te follows any letter and the MimQ none means that the Mim which comes after the Te is not connected following to any
32. ation as K2 The following chapters include Chapter 2 Related and previous work on this subject Chapter 3 Rules and Instructions for Persian Calligraphy and Nastaliq Chapter 4 Research leading to this thesis Chapter 5 Modeling the Stretching Chapter 6 The Nastaliq Font and Chapter 7 Future Work Chapter 2 Related and Previous Work 2 1 Formatting Bi Directional Text There are several programs that handle both right to left and left to right typesetting We restrict our attention to programs that handle Arabic Persian and related languages Almost any program that can typeset Arabic can also typeset Persian The following subsections describe programs for general bidirectional typesetting or for typesetting Persian 2 1 1 Ditroff ffortid The bidirectional version of ditroff ditroff ffortid was built in a modular manner by adding a post processor ffortid to an unchanged ditroff 2 ffortid 1s responsible for printing right to left text from right to left while ditroff treats all text as if it were written from left to right Because ditroff was not modified at all all ditroff pre processors and macro packages work for ditroff ffortid Moreover since ffortid output looks like ditroff output all ditroff postprocessors work for ditroff ffortid Johny Srouji and Daniel Berry describe an Arabic extension to ditroff ffortid that can handle much of Persian as well 5 The algorithm they use for reversing left to right
33. chable Letter Pe an Easily Decidable 6 Point Stretchable Glyph N B a A P ra Figure 46 A 4 Point Stretchable Letter Fe an Easily Decidable 4 Point Stretchable Glyph 74 Figure 47 Letter lt Kaf a Glyph with Misplaced Start Points of Curves Shown in Larger Circles 75 SA Kafs5 Figure 48 Stretched Stand Alone Version of Letter lt Kaf introduced in Figure 41 with Curves A and C Stretched SA Kaf5 Figure 49 Stretched Stand Alone Version of Letter lt Kaf introduced in Figure 41 with Curves B and D Stretched 76 Chapter 7 Conclusions and Future Work 7 1 Summary The research described in this thesis has focused on the problem of stretching for Persian calligraphic typography The first intention was to implement a dynamic font which would work with ditroff ffortid modified as necessary to allow formatting of calligraphic Persian language However after attempting to specify the software for the task it became clear that due to the complexity of Nastaliq it would be impossible to implement its stretching in a dynamic font So the goal of the research was changed to provide a PostScript font which would contain all the allowed variations of stretching of each stretchable glyph so the font would be usable on its own Several TrueType Nastaliq fonts were converted to PostScript to find out which one would include all the features I needed I ended up working with a free font called IranNa
34. d this responsibility for the judgment of the choosing what is more beautiful is given to the author In order to make these rules applicable I had to make a set of strict rules and therefore I simplified them to make them implementable in a software application 3 2 Rules These rules are derived from several instructions and tutorials 12 13 14 15 16 Rule 1 For regular lines up to 3 stretchings in a line of text is always acceptable More than 3 stretchings is possible at the writer s discretion to help keep balance in the line Rule 2 It is not desirable to have two adjacent stretched letters 22 Rule 3 The space between two adjacent letters that are not connected should be one dot Figure 17 shows the Stand Alone letter Be which contains a typical dot which is identified by a rectangle SA Be Figure 17 Stand Alone Letter Be The Typical Dot is Shown in the Rectangle Rule 4 There are three reasons to stretch letters of which only the third is important for typesetting software since the each of the first two needs a human to choose what is more beautiful and the second one needs a human to distinguish the letters by the meanings of the whole words containing them 23 I look more beautiful I avoid confusing between similar words and III justify lines and fill spaces Rule 5 Different amounts of stretching are allowed by different master calligraphers We adopt the convention that a
35. e of left to right characters in the line fi od 2 1 2 TEX XET The most ambitious TpEX based project aimed at formatting Arabic is the TEX XpgT program developed by Pierre MacKay by modifying TX itself to be bidirectional 6 The program does the same reversal of text in designated right to left fonts that ffortid does but inside the modified TX using the internal data structures of the program rather than the dvi output It assumes separate letters but does nothing about stretching 9 2 1 3 Microsoft Word The 2002 Microsoft Office system provided a right to left functionality and features for entering editing and displaying right to left or combined right to left and left to right text 7 Microsoft Word 2007 provides support for many more languages and has solved many of the bugs On Windows XP MirEmad 8 allows Microsoft Word to typeset Nastaliq making Microsoft Word one of the few applications so far handling typesetting Nastaliq and bi directional formatting at the same time However the amount of stretch is constant and there is just one stretched version of each letter 2 1 4 Persian Only Programs There are programs that can typeset Persian with several fonts However these applications do not also typeset left to right languages Therefore they are not bi directional NameNegar 9 and Maryam 10 are examples of this kind of program 2 2 Keshideh In Persian stretching is commonly known as keshideh Kesh
36. ed by the stretch amount horizontally but the end point does not move There are several cases to consider For example what happens if a glyph contains more than one closed paths How does the program know which path is the main path What happens to the other closed paths after stretching To write this code I had to come up with a profile for different characters It is amazing that almost every character has its own properties so I need to consider all of them in order to let the user handle them This issue is explained completely after the 6 point glyphs are introduced 63 Figure 37 A 6 Point Stretchable Curve Glyph Stretching a 6 point glyph is the same as stretching a 4 point glyph There are more cases that need to be considered SD UE ae pe Ne eo 10 11 12 13 14 15 16 Call the current point Cp If Cp lt X10r Cp lt X2 then don t move the curve or line starting Cp Else If Cp X or Cp Xz then Call the end point of the curve starting at Cp Ep If Ep lt X1 0r lt X2 then don t move the curve or line starting Cp If Ep Xsor Ep Xethen stretch 3 Else if Cp gt X3 or Cp X4then move the curve or line starting Cp by the stretch amount horizontally Else If Cp Xsor Cp Xz then Call the end point of the curve starting at Cp Ep If Ep gt X30r gt X4 then move the curve or line starting Cp by the stretch amount horizontally If Ep Xsor Ep Xe then stretch 4 Else if Cp lt X3 or Cp X2 and
37. eme ec M tease trom eases dte opda cn onto 78 Jo Future Work and Exists PRODICINS cresci E 79 7 4 Breaking a B zier curve into Two B zier Curves Given its four Points and a PomtPliced on the Nik chee anne ot enn ne ned taret dee dia uda qan eds 80 7 4 1 Combining Two B zier Curves in Order to Get a New Curve 81 BIDHO TIDI y paenmerreeen errr menor vente Corre adito d yy trance Serene E 82 Appendix A NAS Troff User s Manual ccccccccccccccceceeeeessssssesseeseeeeeeeeeees 84 Vill List of Figures Figure 1 Different shapes of letter Be in different words shown within in the red rectangles A stand alone B connect previous C connect both and D connect following 1 Figure 2 A sample of Chinese Text Each Letter Has a Vertical Movement From Its Previous One Om The B3se ID secca iet Debet S ER a risate P tatiana reda pe DE 3 Figure 3 A sample of Hebrew Text Each Letter Has a Vertical Movement From Its Previous One on 1 Ne Base une stunden uut aee bu decet iem dessa bu tende cen Uu e Lu doest 3 Figure 4 A Sample of Two Words in Persian Written in Book Typing Typeface on the Left Side and Nastaliq on the Right Side i eecccseccccceceeeeseeeeeeeeeeeaeeeeseeeeeeeeeeaeeeeeees 4 Figure 5 Different Examples of Stretching in A Hebrew B Arabic Naskh and C Nastali coetui to Mob ntt MOS nda dut adde NET alert i Meets beg asian 5 Figure 6 Stretching a Four Point Curve Using the Method of Dan
38. f the keyboard This can be easily done be clicking on the language bar at the top left side of the screen and choose Persian ISIRI 2901 from the menu If this language does not exist in the menu simply click on the system preferences on the dock at the bottom of the page and choose international Input Menu Persian ISIRI 2901 Below you can see a sample Troff file handling Persian Nastaliq Here is a sample set of Troff command you need to start typing Persian F PN marks the beginning of using Nastaliq font fP Marks its end OA means other abstract header lp means left adjusted paragraph OA Ip F PN S amp S eu pw yl ob cue ULeiol Sly cul Jlo So cul Uli aol f Hditroff ffortid fPN pl y 625 S jl lo aoU p cul 50 Y 89 Textedit fPN sld exu y aS V Apple Mac OS X PN puiSso oslaiwl 5979 jl MP It is severely hard to show the time sequence in the editor used to type this manual for a text containing both right to left and left to right languages So it is just enough to remind that Farsi unlike English is right to left a PA e Y P a Ld die 3J Textedit o gno ditroff ffortid I a ag ae todo ae P vd siu Apple Mac Os X Figure 3 Sample Persian Nastaliq editing with Troff 2 Chapter 2 Conventions 2 1 User assumption 2 1 1 The User of NAS is assumed to have a good knowledge of Troff If not please refer to the user manual of Troff Plan 9 5 2 1 2 Na
39. ffortid output to PostScript 93 Unicode A standard encoding allowing computer based systems to consistently represent and manipulate almost any existing character in many language or formula There are several versions of Unicode as follows UTF 6 uses 1 byte for all ASCII characters which have the same code values as in the standard ASCII encoding and up to 4 bytes for other characters UCS 2 uses 2 bytes for all characters but does not include every character in the Unicode standard UTF 16 extends UCS 2 using 4 bytes to encode characters missing from UCS 2 Textedit The default text editor on Mac OS X Systems which uses Unicode with UTF 8 and UTF 16 MAC OS X The current 2009 version of operating system implemented on Apple Macintosh machines 94 2 4 Rules in Persian Nastalig This section just contains the actual rules in Persian Nastalig Chapter 4 is about what I will and can implement Rules needed to be taken care of Rule I Page 23 Rules of Calligraphy by Amirkhani Adaabol khatte Amikhani For regular lines up to 3 stretches can be still acceptable leaving the decision for the author and can help to keep the balance of the text Stretched letter should not be selected from the beginning of each line It 1s also not desirable to have stretched letters beside each other Rule 2 Page 9 Learning Nastaliq selected parts of secrets of Nastaliq For not connected letters which come
40. ger and smaller circle 71 BoxCalculator Stretch StretchGlyph4 StretchGlyph6 This program goes over each glyph in the font and calculates its font bounding box coordinates These coordinates are needed when assigning memory to each glyph Also this program finds the FontBBox which is the minimal bounding box which can contain all the glyphs This program gets the control points of a B zier curve and the amount of stretch size of the curve as input and returns the control points of the stretched curve This program gets the PostScript code of a glyph together with the coordinates of the four points of the stretchable B zier curve identifies two stretchable curves in the glyphs and stretches the glyphs using the Class Stretch and returns the code for the stretched glyph This program gets the PostScript code of a glyph together with the coordinates of the six control points of the stretchable B zier curve identifies four stretchable curves in the glyphs and stretches the glyphs using Class Stretch and returns the code for the stretched glyph Table 1 List of All Programs Developed to Make the PostScript Font Containing All Stretched Glyphs In the end the font had 1247 glyphs which look very good to any Persian reader I added 369 glyphs to the 878 glyphs that were already in the font 72 6 10 The Problem of Misplaced Curves One other problem that came up in this research was stretching any glyph that did no
41. has a vertical movement from its neighbors in the same column Figure 3 shows a sample of a Hebrew text Hebrew language is written on horizontal baselines the same as English but is from right to left As can be seen in this figure each letter has only a horizontal movement from its neighbors on the same baseline Figure 4 illustrates an example of two words written in Book Typing typeface on the left and their Nastaliq versions As shown in the figure the Book type typeface is written on a horizontal baseline However for the Nastaliq versions the letters have both vertical and horizontal movement The rectangles in this figure show the first letters in each word and the circles show the last letter in the words As can be easily seen each rectangle is positioned both vertically and horizontally different for the circle in the same letter which means there exists both horizontal and vertical movements 2 E d kK R EXE Hh 1 rn x HH m ok d He 3E E BRK FTF X KR X cK 15 ff KF HF i X m EF M m gy Zk m X Kf Hh gd o m amp d T d ht f a H g KR HT o M W Hog A mM o dmm K Z lt sg X Rk hh dx MS SU f s RH KH X th d i nn A 8B m A m mx x dd au RK gy EX Figure 2 A sample of Chinese Text Each Letter Has a Vertical Movement From Its Previous One on The Baseline JPYINR 1373 Ty Dey TART m3 ga WY D iyg Ten Wry npw PA Di v npw WIAD inqinny vong c oyq 739 Tina Figure 3 A sample of Hebrew Text Each Letter
42. having many stretched letters A half stretched letter is a letter which is not stretched by the maximum permitted amount Sometimes in order to make achieve line alignment we may decide to put more than one stretched letters in it Rule 12 It is important to check each stretchable letter s location in its line before stretching it When there is more than one stretchable letter in a line it is required that there be some distance between every two stretched letters To be more specific divide each line into 5 equal width sections from right to left If one letter ends up in two sections it must be considered as a member of the section that contains a larger portion 26 of the letter If the letter is equally in two sections it can be considered to be in either section If there is to be one stretched letter in a line the best place for it to be is Section 3 If no letter in Section 3 is stretchable the next candidate is Section 4 and if no letter in Section 4 is stretchable the next candidates are Sections 2 5 and 1 in that order If there are to be 2 stretched letters in a line the preferred places are Sections 3 and 5 If neither section has a stretchable letter or they have stretchable letters but some rules might be violated by stretching these letters the next candidates are the couples of Sections 2 and 5 Sections 2 and 4 and Sections 4 and 1 in that order 3 3 How to stretch a letter in Persian Nastaliq As discussed
43. hown within the Rectangle 30 Figure 21 Nib Head Using the Power Style and Solid Mode in Letter Te 31 Figure 22 Nib Head Using Power Style and Rotate Mode in Letter z He 3l Figure 23 Surface Rased with Edge l Shown in Grey cccccccsssseeeeeeeeeeeeeeeeeeeees 37 Foure 24 Curves ot Type T rl aet EE RR Ens 38 Pieutre 29 Curves Ol Type 2 2 s ie e POR ii eh iet Um 38 Figure 26 Stretching a Curve Belonging to the Set 81 with Ebe B1 B2 40 Figure 27 Stretching a Curve Belonging to the Set 2 with Eaf B1 B2 4 IX Figure 28 Connect Following Letter 4 Ghaf Followed by One of the Letters z Jie ehehe He UN MC hira EETT 53 Figure 29 Connect Following Letter 6 Ghaf Followed by Letter e Mim Which is Not Followed Dy AVVISI sore ose eut wat eduey ete sedato dus cba doa ead evade dedu EE doe 53 Figure 30 Connect Following Letter 3 Ghaf Followed by Letter e Mim Which is Followed by Another ettet rescues stun Feud vt dese eod setne usta dou Ret uti rue tdt 54 Figure 31 Different Combinations of Letter Ghaf cece ccccceeeeteeeteeeeeeeeees 54 Figure 32 Random Naming for Glyphs in the Original Font Letter s Ye 23 Figure 33 A Sample Naming for Stretched connecting previous Pe 56 Figure 34 Non Stretched Letter Be Which is Followed by One of the Letters Be Pe Te or Se Which is Also Followed by One of the Le
44. ideh has different types as explained next 10 2 2 1 Different Versions of Stretching 2 2 1 1 Stretching the Letters This case occurs when we stretch the stand alone and connecting previous version of the letters This case happens for letters Be Pe 5 Te Se o Sin Shin Fe S Kaf and S Gaf 2 2 1 2 Stretching the Connections This case occurs when we have connected letters in a word In this case respecting the rules introduced in Chapter 3 we can stretch at most one connection in a word 2 3 Pre complied Variations and Dynamic Stretching There are two ways considered in this research to implement stretching First in the dynamic fonts each glyph is described by a parameterized procedure that draws the glyph s outline in a variation determined by the values of the parameters passed to the procedure Notice that there can be an infinite number of versions for each glyph since each parameter can be any value This technique is called dynamic stretching The other solution is to construct a finite set of stretched versions for each letter In this case we can design all the stretched amounts of the letter and select the correct glyph 2 4 Dynamic Font for stretching letters Daniel Berry has implemented stretching of letters in Arabic Persian and Hebrew 2 Although the Persian typeface he uses is different from Nastalig many parts of this work are 11 useful for stretching Nastaliq The
45. iel Berry 2 12 Figure 7 Cornered Shared End Point Caused by Stretching seuuuss 13 Fic re o FIXNE ie COMER senio t Et e desc use NEL anes ec ce cate 14 Figure 9 Special Case Horizontal Tangent eese 15 Figure 10 Various Attempts to Stretch the Stand Alone Alif Maksura 16 Figure 11 An Example of Different Uses of the Nib Head in One Letter 17 Figure 12 a Sample Nib Head b The Stand Alone gt Reh Using the Nib Head c The Written Part of the Letter gt Reh d The Drawn Part of the Letter gt Reh 18 Figure 13 Unstretched Stand Alone 3 Qaf nnssssssssseenssssssssseersssssssssseersssssssseeeees 19 Figure 14 Long Stretching of the Stand Alone 3 Qaf in Daniel Berry s Method 19 Figure 15 Nip Head am Naskh 5016299 0 etie tus a E 20 Figure 16 A Sample of Correct Stretching in Arabic Naskh Respecting the Constant THICKNESS of the NID Hedd suia icri n tae berba Et a dud teo habi pax boo ev tt onu pe cbe tuus 21 Figure 17 Stand Alone Letter Be The Typical Dot is Shown in the Rectangle 23 Figure 18 Unstretched Stand Alone Letter Be with Size 5 Dots The Typical Dot Wan be Seen in the Bottommor the Weer sie ica teres chester iia ees onde erasa N 28 Figure 19 Stretched letter Be of Figure 18 by the Amount of 4 Dots 29 Figure 20 Use of the Weak Style for Letter Jim S
46. in the ends of most of the letters like Be 2 Re Sin and c He in which not all of the nib head is placed on the paper Only part of the nib is used to draw the weak part of the letter The rectangle in Figure 20 show the usage of nib head in weak mode As shown in Figure 20 the nib head is using the weak style everywhere inside the larger rectangle at top of the figure In this area the nib head is larger than the drawn letter s width which means that not the whole head is used to draw some part of the letter To achieve using just some part of the nib head one places enough of the corner of the nib head on the paper and moves it on the paper The smaller rectangles stands for the nib head 29 7 Figure 20 Use of the Weak Style for Letter z Jim Shown within the Rectangle The second style is called the power style in which the entire nib head is placed on the paper There are also two modes for the angle of the nib head for this style In solid mode which is used in letters such as lt Kaf lt Gaf Be 9 Pe Te amp Se and Fe the nib head s angle does not change and stays almost the same in order to write the letters However in practice it may rotate by 20 degrees Figure 21 shows a version of letter Te stretched by the amount of 3 dots The section within the larger rectangle 1s drawn using the solid mode The other mode is called rotate mode in which the nib head rotates as in letters
47. ine formatting for Arabic Hebrew and Persian with ditroff ffortid and dynamic POSTSCRIPT Fonts Software Practice and Experience 29 15 1417 1457 3 Bayar A amp Sami K January 2009 How a Font Can Respect Basic Rules of Arabic Calligraphy International Arab Journal of e Technology Vol I No 1 4 Andre J amp Borghi B 1989 October Raster Imaging and Digital Typography J Andre amp R D Hersch Eds Proceedings of the International Conference Ecole Polytechnique Federale Lausanne pp 199 207 5 Srouji J amp Berry D M 1992 Arabic Formatting with ditroff ffortid Electronic Publishing 5 4 163 208 6 Knuth D amp MacKay P 1987 Mixing Right to left Texts with Left to right Texts TUGboat 6 1 14 25 7 Right to Left Language features Help and How to Microsoft Office Online n d Retrieved from http office microsoft com en us help HA 10351021 1033 aspx 8 Maryam Soft n d Retrieved from http www maryamsoft com Default asp Page Miremad htm 9 Nameh Negar n d Retrieved from http aramedia com namehnegar htm 10 MaryamSoft n d Retrieved from http www maryamsoft com Default asp Page Maryam htm 11 Fazaeil H 1987 Learning Nastaliq Tehran Soroush 12 http www bluej org n d Retrieved from BlueJ Teaching Java Learning Java 13 Amirkhani G 1982 Amirkhani s Calligraphy Tehran The organization for Calligraphy i
48. information that is useless for Type 3 fonts In order to be able to see what I was doing I converted the Type 3 font into a PostScript program that displayed all its defined glyphs 6 5 Glyph Identification I was able to see that the initial font contained about 1366 glyphs which had nothing to do with Nastaliq I had to remove all the unnecessary non Persian letter glyphs I mentioned in the first chapter that Persian has 32 letters and at most 4 positions per letter in normal typography Therefore there should not be more than 128 glyphs After removing unnecessary glyphs there remained 878 glyphs from the original 2244 glyphs Thus in Nastaliq every letter has about 27 different glyphs Some of these shapes were so similar to each other that it was hard to determine if all but one of a set of similar glyphs could be thrown out or each similar but not identified glyph served different purpose Figure 28 29 and 30 show the letter 3 in three very similar shapes The first shape is when amp Ghaf is connected following to e Mim while e Mim itself is not connected to anything else The second shape is when the amp Ghaf is connected to following e Mim and the following after e Mim is connected to another letter and the third shape is used when amp ghaf is connected following to any of c Jim Che z He and e Khe Figure 31 shows letter 3 Ghaf in the three combinations described in the previous three figures Next
49. int M55 They conjectured that the new curve with its four B zier control points respects the spacing issue of respecting the constant size of the nib head between the curves in letters or in other words if we stretch the curves in the original glyph using the transformations they have provided the resulting glyph containing the new stretched curves would have the same thickness as the original one The proof of this conjecture was not supplied by Abdelouahad Bayar and Khalid Sami So I provide this proof as a part of the research for this thesis Mona Mojdeh has also derived an independent proof that is not presented here To prove this claim assume there are two curves spaced from each other along the line having a 70 degree angle from the baseline It is necessary show that after applying the formula the resulting curve s control points are spaced the same distance Step 1 simplifying the formula Assume that current curves are made with control points Mj 9 M44 M45 M5 and M5o M21 M22 M55 Denote the variables for the shifted version of the curve with a t notation at superscripted at the top of its metric variable So M Mj M gt Mj l M Mi M23 Mi By considering just the x value of each Mj we get X117 1 Ci1 X10 Cilx X12 1 C2 lx C2X13 X11 e A10 i ey Pn or 43 X12 1y EN COTES Mii 1 557129 Qo h Yao ls ty Miz7 x Ux by e X13 Y13 Step 2 Calculating for the
50. itions Note The slanted line drawn on the stretched letter means that it is a mistake to stretch it coe 5 i sg Before as in Dom ry Before another stretched same letters as in ot i Before as in Before as in Before U as in Tai Rule 10 Page 14 A look into cheminstry in Nastaliq 3 In every line it 1s better to have 1 stretched letter or 2 half stretched ones Sometimes in order to make the line look better we may decide to put more than one stretched letters in 97 every line It is better to not decide the stretched letters in lines so that they do not locate beneath each other This helps the text to not look more crowded in some parts of a page than other parts For a letter even it 1s stretchable it is important to check its location in the line before stretching it due to calligraphy rules in Persian It is important that in every line we consider at least one stretched letter and with respect to the kind of the letters we can put more stretches in each line however it is required that we consider some space between every two stretched letter To be more specific let s divide each line into 5 sections as shown below We can divide every line to five sections just considering the line not caring about the words in it Also if one letter 1s in two sections when we divide the text into 5 sections it is possible that a dividing line divides a letter into two sections In the following figure
51. l curve of degree at most 3 can be expressed as a B zier curve 10 by breaking this polynomial into two polynomial curves of the same degree we can get two B zier curves This process also needs direct human intervention to choose the right break point on the glyph s layout to achieve a more beautiful glyph after stretching 80 7 4 1 Combining Two Bezier Curves in Order to Get a New Curve As a matter of fact it 1s not possible to combine two B zier curves of the same degree and get a single B zier curve covering the same points with the same degree in general But there are special cases in which this can be done For instance if the starting curve is the result of the splitting operation descried in Section 7 4 it would be possible to combine the curves More generally we can divide every curve into many break points Assume we have n break points K K2 K3 Kn The trick is to calculate a polynomial P n of degree 3 which would minimize the sum of the square distances between P n and every Ki for i 1 2 n This method is called the least squares method and is applied extensively in optimization models The resulting curve might not be exactly what we want but it might still be useful improving the appearance of stretched glyphs 8l Bibliography 1 Persian language Wikipedia n d Retrieved from http en wikipedia org wiki Persian_language 2 Berry D M Stretching letter and slanted basel
52. ld best be useful for my purpose As mentioned in Section 6 2 the IranNastaliq font was the one chosen to proceed with As explained in the next section I had to stop and rescope the whole idea of my research before starting to read the code for ffortid I learned the importance of writing some requirements document before starting to code In this case the user s manual served as my requirements document If I had started coding before preparing the manual I would have wasted a lot of time writing and changing code before I ended up learning that it was 1mpossible to reach my goal Therefore spending about 6 months of working on gathering information for and preparing the user manual saved me about one year of reading the entire ditroff ffortid code getting familiar with any already existing processors handling the features that I am interested in and writing new classes and methods 34 4 4 Rescoping the Research In the middle of writing the manual I realized that the goal could not be achieved in a time reasonable for a Masters thesis I learned that stretching could not be achieved entirely by use of a dynamic font The reason was that human s interaction was necessary for many glyphs in order to recognize which part of the glyph and which B zier curves had to be stretched Therefore hundreds of particular instances of stretched letters would have to be pre compiled into the font As a result I with my supervisor s approval resco
53. lected parts of secrets of Nastaliq Mahmoodi Seyyed Mahdi ISBN 964 90656 6 0 Publisher Moallef 6 Edition Unicode and Multilingual Editors and Word Processors for Mac OS X http alanwood net unicode utilites editors macosx html 103
54. letter 56 CF Be OBe Pe TelSeCQOBe Pe Te SeC 0 Figure 34 Non Stretched Letter Be Which is Followed by One of the Letters Be Pe Te or amp Se Which is Also Followed by One of the Letters Be Pe Te or amp Se CB All Te MimQNone dd p Figure 35 Non Stretched Letter gt Te Which is Followed by a Letter Mim Which is Not Connected Following to Any Other Letters 57 The naming process contains the following Grammar lis for or O and C are for grouping O is and C is ix means x is for zero or more occurrences of what is applied to is empty string Nonterminals are in Comic Sans Ms Terminals are Calibri There are no spaces in the strings generated by this grammar spacing in these rules is for the readability of the grammar GlyphName Connectivity ContextPrevious JGlyph ContextFollowing Connectivity SA CP CF CB Glyph Letter Length O Letter Length c Length NumeralAtLeastOne K NumeralAtLeastOne 1 213 4 9 Letter UpperCaseLetter LowerCaseLetter OLowerCaseLetterc Digit20 1 2 31l4 l9 UpperCaseLetter D A Bl c bpl vlz LowerCaseLetter Dalb lcl dl vlz 58 ContextPrevious Context ContextFollowing Context Context Any Rest ListOfChoices Null ListOfChoices Choice ORSign OChoiceC Choice LettterOrNone OQ LetterOrNonec LetterOrNone Let
55. m translated the original names of each glyph in the font to its new name which encodes the purpose of the glyph For this purpose I had to implement a parser that parses the font file and finds each instance of a glyph name while leaving everything else untouched Abdelouahad Bayar and Khalid Sami s formula had to be implemented to generate the stretched glyphs from the existing ones This program is given the control points of each curve as input and outputs the control points of the stretched glyph For this purpose two conditions have to be taken care of First the start point of the curve which is going to be stretched must be is on the right side of the end point and second the start point must be on at the left side of the end point I refer to curves of the first kind as Type 1 and the second kind as Type 2 Here is the algorithm my program uses to stretch a glyph As mentioned previously each letter can be stretched minimum by the amount of from one to five dots Let call the file containing the glyphs PostScript code fontFile Ps Read from fontFile Ps and get condition and number of control points from the user If there are four points this means two points are located at the left side and two are located at the right side as shown in Figure 36 Xz X2 are left and X3 and X4 are right 61 If a current point is at the left side of X and X2 leave the curve or line starting at current point untouched If the current point
56. mes and insisted on preparing the user s manual first I decided to start with working on three issues e The rules for Persian Nastaliq e The available Nastaliq fonts e Reading and understanding the code for ffortid First in order to prepare the rules for Nastaliq I had to find reliable sources For this purpose I took advantage of a trip I had made to Iran I met Mr Majid Hosseinzadeh the Head of the Association of Calligraphers in Iran and explained to him what my research is about and 33 asked him for the best available resources to find and learn rules for Nastaliq He suggested that I buy a book written many years ago which is known as the bible of Nastaliq for calligraphers 11 My problem was that this book was no longer sold in public stores for some political reason At last I found the book in Tehran s University Library and copied all the pages of that book I also found and purchased all the other available books for learning Nastaliq in Iran The next stage was to find out if any other software applications handled the features of Nastaliq that I was studying The next stage was to find and purchase all the available Nastaliq typesetting software applications By installing and using all of them I found out that none of them could handle the features I wanted to implement The next step was to read the references I found and gather all the rules which would affect my software I had also to find out which available font wou
57. n a T Y 13 l sin a Y 12 l sin a SIL dU e CIEL OUS dL abt Ssmo ED co tlg l cos a 13 sinia v X 11 cos a X149 l cos a Y11 l sin a Y40 1 sin a Y13 l sin a Y12 1 sin a X11 1 cos a X 10 l cos a X13 l cos a X 12 1 cos a X 1 cos a And the resulting Yt will be 46 Y11 Y10 Y 43 712 X190 h Y419 X13 T Y43 v y 117 0 REN wW 10 X13 X12 Remember EA VNCHNS OLI ED X419 th Yio Be X11 X10 X13 X12 _ Y31 Y10 E E um Xj Xio h Yo Xj 1 cos a X19 l cos a h Yo n ee Y11 sin a Y19 l sin a His X11 cos a X19 cos a sin a Yz l sin a The new X will be Xr PIS ee oe X12 l cos a ea TTL on Yio l sin a Y12 l sin a X10 l cos a X11 L cos a X 19 l cos a Y11 Y10 _ Y13 Y12 X117 Lcos a X19 Lcos a X13 Lcos a X 12 l cos a Xr l cos a The new Y will be Y11 Y 10 Y 13 Y 12 X10 X42 YVi0 Y12 X11 X10 X 13 X 12 ERA X45 598 Y11 Y10 Remember Y ZES ZTS ZES E 11 Y10 _Y13 Y12 X11 X10 X13 X12 Yi1 Y omar X19 Yi X l cos a X19 l cos a Yio 0 sin a 11 L sin a Y49 l sin a E q X11 l cos a X 49 l cos a gt Y l sin a Now Remember Mt 9 X10t I cos a Yiotl sin a Mt 1 X114 l cos a Yi sin a
58. n Iran 82 14 Amirkhani G 1997 Rules in Nastaliq Tehran The Organization for Calligraphy in Iran 15 Falsafi A A 1999 A look into combinations in Nastaliq Tehran Yasavoli 16 Mahmoodi S M 2000 Nastalig s Complete Tutorial Tehran Moallef 17 EL Houssaini A n d The Arabic Caligraphy in Arabic 1994 18 Essaid Mahmoud M 1994 Learning Arabic Calligraphy Naskh Roqaa Farsi Thuluth Diwany Cairo Egypt Ibn Sina 19 El Khattat M H 1986 Arabic Calligraphy Rules A calligraphic set of Arabic Calligraphy Styles Beyrouth Lebonan Book Univers 83 Appendix A NAS Troff User s Manual 84 University of Waterloo Shahab Mohsen NAS TROFF USER S MANUAL This Document contains the User Manual for NAS a modification to ffortid to let Troff support the Persian calligraphy called Nastaliq and also support its stretching mode 85 Table of Contents MEM ree O RENE 87 Ld CPEIDOUCDO VELO IC Wo a a Uo soseqcte IOS enwneees Ceu DENN Bent I Rd eensmee cena eos 87 NAME eM 89 2 Er CONN EIE 90 a UTT O E E EA 90 2 1 1 The User of NAS is assumed to have a good knowledge of Troff If not please refer to the user manual of Troff Plan 9 5 sss 90 2 1 2 Nastaliq has many different rules used by different Masters in Iran Here we obey mainly the rules from the sources we mention in the bibliography 1 2 3 4 90 2 2 National Gy OG pratense tsutqs HE UeuNoNt
59. n its word These shapes are mainly known as Stand Alone Connecting Previous Connecting Following and Connecting Both Some letters have fewer than four shapes Figure Ishows all the four different shapes for letter 2 Be in different words The fact that every word can have up to four different shapes makes Persian typography a big challenge A B C D Figure 1 Different shapes of letter Be in different words shown within in the red rectangles A stand alone B connect previous C connect both and D connect following Similar to other languages Persian also has several typefaces The regular Book Typing typeface which is the type face already used in Figure 1 1s written on a horizontal line Other typefaces may differ 1 2 Persian Calligraphy The best known Persian typeface which is also a kind of calligraphy is Nastaliq Nastaliq has some properties which differentiate it from other typefaces Each letter has both a vertical and a horizontal shift from its neighbours in most other languages each letter is positioned either horizontally or vertically with respect to the previous one Even in regular Persian book typefaces we just see a horizontal shift Figure 2 illustrates an instance of a Chinese Text To write Chinese traditionally letters are arranged in vertical columns read from top to bottom down a column and right to left across columns As can be seen in this figure each letter only
60. nd name each glyph as a positioned letter in a particular context This thesis describes all the steps taken to build the font iii Acknowledgements Thanks are given to my supervisor professor Daniel M Berry whose help made this research possible even with the existence of many issues I also would like to thank my readers Professor Richard Trefler and Professor Craig Kaplan I also thank Mr Hosseinzadeh the Head of the Association of Calligraphers in Iran whose kind advice helped me to get access to the best resources available 1V Dedication Hereby I dedicate my thesis to my lovely wife and to my beloved parents who have Supported me in every situation in my life Table of Contents AUTHORSJDECIZXRATIOJN teet ion tiet Rh nhe Put meas ii PROS UAC met ETE iii ACKNOW IEG S CMICINS ERU UT UU UT iv Dedi HIT DR RT T LO OO OD LEE V TabIe DP C OnDte niis docs ecd S E tT M du vi Etot T T TI UU IU U T 1X Chapter IntroduetlOl aimee tera a a danni eaten eta I 1 L Persian Wrin oreste tire etd se Eurus OF etd sn eA mU as osdta cub Fuge odis Ebo l 1 2 Persian Calera ea tia ta aW EROR ax vimus aS this peus d USE 2 1 3 Problems with Automating Persian Writing and Persian Calligraphy 5 4 Notalional ONY CMUONG sucio atio epe oat S 6 Chapter 2 Related and Previous Work cccccccccssssseeeeecceeeeeaeeeeseeecceeeeeaeeeeeees 8 2 1 Formatting Dr Dir cttonal Lext sonde evt peu ae EE eas Len tea cer are abus 8
61. of stretch is in the middle of one cubic B zier segment 13 TA an i fi i a PA NS B A RB prr ax Am F Figure 8 Fixing the Corner One special case of stretching through a shared end point works specifically when the tangents through the shared end points are completely horizontal as suggested by Figure 9 14 4x ai a i i i i X amp 7 a a 7 2 A T Figure 9 Special Case Horizontal Tangents 2 5 Problem with Daniel Berry s Solution Although this kind of stretching works and provides a stretched letters it still has some problems As can be seen in Figure 10 Daniel Berry provides some various attempts to stretch the Stand Alone Alif Maksura in Arabic which is the same as the letter Ye in Persian 15 GOGG GGG G KULLU Figure 10 Various Attempts to Stretch the Stand Alone Alif Maksura Daniel Berry has manually made some different version of letter Alif Maksura in Arabic As can be seen in Figure 10 the thickness of the horizontal part of the letter Alif Maksura shown in the right side of the figure differs in all variations These different looking versions of stretching are symptoms of a single fundamental problem described in the next section namely 16 that Daniel Berry s implementation of stretching does not model the way stretching is done in hand written Persian or Arabic writing 2 6 Abdelouahad Bayar and Khalid Sami
62. olute values for coordinates when converted to PostScript unlike the other two fonts It is easier to understand absolute coordinates and see how to modify them Also as mentioned in the next section the conversion failed for the NamehNegar font 6 3 Conversion Tools I needed to convert the TrueType IranNastaliq font to a PostScript font To convert this font I used a program called ttf2ptIthat converted its input TrueType ttf font to a PostScript Type 1 font together with a afm file which contains standard font metrics The ttf 2ptl program was applied to all three fonts mentioned above The result was that only IranNastaliq and WM Nastaliq were successfully converted The resulting font metrics file gives only horizontal movements which is for each character the distance to the next For normal Latin fonts only the horizontal movement is needed since all lines of text are horizontal In a font for slanted baseline writing the vertical movement of the next character might also be needed One big problem was that no font conversion program would generate the vertical movement for the metrics file Even FontForge which is one of the most well known applications for font conversion in Mac OS X does not compute vertical movements 6 4 Converting t1a to Type 3 It was necessary then to convert each Typel font to a Type3 font This conversion required changing the commands and getting rid of the hinting commands which provide 51
63. or The next state is ptrn s turn which would do letter form and ligature identification on letter only input to yield output with each glyph to be printed Also shape changes which may occur are handled in this section The result which is Unicode together with some ligatures will be delivered to ditroff ffortid Now with the changes implied to the ffortid class the software calculates where to places each letter and also how much each letter should be stretched The last step before the output 1s generating the post script code first and then applying the change in shapes which will be done to the post script code of each letter Figure 2 also provides a big picture of the flow of data in the software Persian Translator idt ausis Position Ligature ext wath Identification and Identification ligatures stretch calculation Make changes in Translation to shape 1f stretched glyph codes letters Figure 1 The flow of data from keyboard to output result 88 Letters Glyphs ffortid DIF DIF PS Time Order Time Order Visual Order Figure 2 High level view of NAS and troff Processors 1 2 A sample Run Writing Nastaliq can be easily achieved by using NAS First you need to run the Textedit on your Mac OS X NAS not only handles documents only in Nastaliq but it also handles making documents containing both left to right and right to left languages In order to write Nastaliq we first need to change the language o
64. ortant to notice that for the stand alone Letter S Kaf there exists no rule explaining how much the Hamza should be moved and the decision is left for the writer to where to place it horizontally above the letter SA Kat Figure 41 Stand Alone Letter lt Kaf Not stretched 69 SA Kaf9 Figure 42 Stretched Letter lt Kaf For letters Khe z Jim Che amp ghaf Fe 5 Za 4 Ta V Shin and Ghein the dots need to get shifted by the stretch amount N i i CE AN fen d OV B A a os a a p V Figure 43 Connecting Both Letter Fe of the Size 5 Points 70 gt OL o A SS Hg os A pe ST Figure 44 The Connecting Both Letter Fe Stretched Version of Figure 43 by the Amount of 4 Points I used the BlueJ 12 programming environment to implement this software Table 1 presents a list of the programs and their descriptions Parse This file gets a font and a list of useful glyph names and parses the fonts MEE PS code and drives the code for the mentioned glyphs Update Update gets a list of new names for glyphs as input and updates all the n glyphs names derived from Parse ShowCurves This program goes over the glyphs in the font and identifies the start point of each curve or line in each glyph which a small circle The very first circle which is the start position of drawing the glyph and the second one are respectively shown with a lar
65. ped my thesis to implementing only the font The development of a new ffortid processor was therefore abandoned and the project changed to that of developing a PostScript font with hundreds of glyphs including up to five different amounts of stretching of several glyphs a PostScript font that eventually could be used with a modified ffortid to be written in the future to work with this font The requirements of the new font are derived from the requirements of the modified ffortid Note that it implements many but not all rules of Persian calligraphy and Nastaliq as mentioned in Section3 1 35 Chapter 5 Modeling the stretching It is necessary to describe the work of Abdelouahad Bayar and Khalid Sami 3 in detail since it is the basis for the research of this thesis Abdelouahad Bayar and Khalid Sami first developed a model of the way the Persian characters are written by hand They determined that the outlines have to follow the path determined by the four edges of the nib head This means that the thickness of the letter along any axis parallel to the angle of the nib head must be equal to the height of the nib head Therefore a mathematical modeling is needed for the software to achieve the correct simulation of Nastaliq handwriting For example In Figure 23 there are two curves B and B2 which is a translation of B1 by vector U 36 Figure 23 Surface Rased with Edge l1 Shown in Grey In Figure 23 B is a B zier cu
66. rs I could trace the path to draw the glyph 6 8 Identification of Stretchable Curves Next I had to identify the stretchable glyphs There are two conditions here 1 The glyph contains 2 curves that can be stretched and 2 the glyph contains four curves that can be stretched Figure 36 and 31 show examples of the two conditions I had to first identify the condition of each glyph I had to examine the code for each glyph identify its stretchable curves and find their start and end points For some of the glyphs each condition should be considered For these glyphs either condition would have worked and I had to choose the one that would look better It is important to mention that this stretching is not dynamic There is a separate glyph in the font for each stretch amount I should notice that the font initially contained at most one stretched sample of stretchable glyphs but these glyphs could not be accessed with an editor and just existed in the font with no use At this stage I learned that the glyph for some specific amounts of stretching do not exist in the original font So I 60 had to make some glyphs manually The process involved copying each part of a glyph from a similar glyph and binding these parts together to get the required glyph 6 9 Implementing the Font To implement this font I developed several Java programs which helped me to do various tasks in order to save my time and simplify the work for me One progra
67. rve with 4 control points Mo M11 M12 and M73 Consider B2 the B zier curve with the four control points Mzo M21 M27 and M23 such that Mz t gt Mii I 012 3 where t is the translation of vector u such that u u u l cos x Ll sin This thesis provides a review on Abdelouahad Bayar and Khalid Sami s work on modeling and supporting keshideh in Arabic calligraphy 5 1 Keshideh the Mathematical Model I use the notation Mo M M5 M3 to denote the B zier curve with control points Mo M M2 M3 Figure 24 shows the set 4 the set of B zier curves Mo M M2 M3 with an invariant concavity verifying 37 M M3 Ai 0 Ang M M i lt Whereiis the axis x director vector and the Ang u V function gives the angle between vectors U V in terms of parameters respecting the positive orientation of the coordinate system Figure 24 Curves of Type 1 fj Figure 25 shows the set p the set of B zier curves Mo M1 M5 M3 with an invariant concavity verifying MoM Ai 0 lt Ang i MM lt gt Figure 25 Curves of Type 2 f 38 Keshideh is a juxtaposition of two B zier curves B and from the set 54 and B2 from the set B5 If Lo Ly Lz Lsare control points of P and Ro R Rz Rs are the control points of Bz then L3 and Ro are equal So the definition of Ej that stretches the curves in 4 and Eg that stretches the curves in f are Ene By X 0 hm X 0 vm gt i
68. s developed in this research to model handwritten Nastaliq iv A Nastaliq Type3 font containing all the stretched version of stretchable glyphs A PostScript Type3 font was developed containing 1247 glyphs including all the possible stretched versions of all possible stretchable letter 7 2 Lessons Learned There were lessons that I learned in this research Most important before starting any software development first prepare a software specification or a user s manual for it The main reason as mentioned earlier is that a good software requirements specification in my case a user s manual is always a good way to discover flaws in one s software earlier I also 78 learned that I had to trust my supervisor because if it wasn t because of his insisting on writing the manual I would have wasted a lot of time on coding without getting any result I also learned that research takes more time than what we expect This is because we do not know all the aspects and therefore at any stage you might find important issues neglected that would need to be considered 7 3 Future Work and Existing Problems A remaining problem is that no available program for working with TrueType fonts can handle fonts with both vertical and horizontal movements As mentioned earlier many convertors can be used to work with TrueType fonts such as ttf2ptl FontForge CrossFont TransType etc None of them could handle horizontal and vertical movement at
69. s Observations Abdelouahad Bayar and Khalid Sami 3 describe the problem that Daniel Berry s method faces Daniel Berry s system offers the possibility of stretching characters horizontally to a degree However in Persian and Arabic calligraphy a vertical stretching might also be required In each of Persian and Arabic calligraphy some parts of letters are written directly with the nib head and other parts are drawn in a way that the contour is set first with the right up corner of the nib head the Qalam s tooth and afterwards it is darkened with a brush 11 12 13 The following three figures show an instance of letter Reh The whole nib head is used for the black part and a corner of the nib head is used for the other part PE ETE Aas er j Figure 11 An Example of Different Uses of the Nib Head in One Letter 17 i a al DJ Figure 12 a Sample Nib Head b The Stand Alone Reh Using the Nib Head c The Written Part of the Letter Reh d The Drawn Part of the Letter Reh The problem with Daniel Berry s method is illustrated in Figure 13 andFigure 14 The thickness of the stretched letter changes and does not respect the constant size of the nib head In some places it is more and in some places the thickness 1s less than the nib head Thus Daniel Berry s method does not model hand written Arabic writing 18 Figure 13 Unstretched Stand Alone o Qaf Figure 14 Long
70. s the suitable letters to stretch using the provided rules in chapter 2 and stretches them automatically by the algorithm and amount explained again in chapter 2 3 1 2 Manual In this way the user has to exit the automatic mode by the na command and enter the Manual mode Then by the strm command before any word the user can ask for the maximum stretch of the last possible letter in the word The user has to use the nstrm command to cancel the manual stretching mode 99 Note Entering the Manual stretching mode would also turn off the automatic stretching mode automatically 4 Design Decision 4 1 Overview In this chapter a set algorithms will be provided which are already implemented in the NAS These algorithms are included in this manual to help the user understand the rules of Nastaliq together with the functionality of NAS better 4 2 Algorithms 4 2 1 Automatic Calculating Calculating the length of existing line The length of the existing line would be the addition of the width of all Maximal connected group of letters plus the space between the not connected letters or Maximal connected group of letters The default size of this space is known to be V2 side of a dot Calculating the available amount for stretching The amount to calculate the free space we have in a line would be equal to Default Length of a line length of the current line When the available amount is calculated we
71. shifted curve Denote again that the variables for the shifted version of the curve with a t notation at superscripted at the top of its metric variable So For Translate every x to x Lcos a and y to y 7 Lsin a So we get Y11 Y Y13 Y X10 l cos a O12 l cos DE a Y10 Y12 X 10 re ie Lo X11 X10 X13 X12 With the same result we can get I sin a i Step 3 Calculating the metrics for the stretched curve Denote that single apostrophe stands for the stretched version Mio M10 h O Mis Mis 0 v Mir 1 Ct Miet Cr Miz 1 C2 J Co Mig Presenting the control points of the shifted curve by vector U cos a sin a with X and Yt we know that Xt11 X1141 cos a X72 X1241 cos a Y Y174 sin a Y Yr sin a 44 X 12 X124 1 cos a X13 X1341 cos a Yt12 Yi2tI sin a Y 15 YisHl sin a M10 X10 Y10 M11 X11 Y11 M12 X12 Y12 M13 X13 Y13 M20 X10F 1 cos a Yiotl sin a M21 X11 1 cos a Y114 sin a M22 X12 I cos a Yiz H sin a M23 X134 l cos a Yis H sin a Step 4 Calculating Metrics for the stretched curves Note again that X means the x value of J and Y means the y value of I Now the stretched version of curve 1 is remind that in the following formulae a single apostrophe stands for a new stretched point M10 X10 Y10 M13 Xis Y10 V _ y Y Y Y The line from Mio to M11 is 2 x X10 X41
72. staliq Due to the complexity of the calligraphy the font contains about 900 glyphs which are used in writing Then each glyph was named by its letter position stretch amount and previous and following contexts This naming process was necessary to make the identification process of each glyph easier It was necessary to prove Abdelouahad Bayar and Khalid Sami s conjecture 3 Also their formulations for Arabic Naskh Stretching had to be changed so that they would work for Persian Nastaliq The last stage was to implement this formulation to achieve all the desired stretched sizes of each glyph There are still some problems left for future work TI 7 1 1 List of Contributions The contributions achieved in this thesis can be summarized to four categories 1 Development of a new regular language for naming Glyphs in a Nastaliq Font As explained in Section 6 6 a regular expression was developed to name each glyph in the font to make it easier for the user to identify it and use it in the right context ii Proof of the validity of Abdelouahad Bayar and Khalid Sami s stretching model Another achievement of this research was proving the validity of Abdelouahad Bayar and Khalid Sami s conjecture about modeling the handwritten Arabic Naskh This proof was necessary to make sure a similar process would be valid for modeling Nastaliq ii1 A stretching model for Nastaliq A stretching formulation based on Abdelouahad Bayar and Khalid Sami s model wa
73. staliq has many different rules used by different Masters in Iran Here we obey mainly the rules from the sources we mention in the bibliography 1 2 3 4 2 2 National Conventions e Times is used for regular text Headings e Times Italic is used for Terms and emphasis 90 e Comic Sans is used for program names and code e Tahoma is used for Sample Input e Century is used for sample output 9 2 3 Terms NAS The name of the program You The person who uses NAS addressed by you Dot A square shape found in some Persian alphabet such as several letters in the alphabet such as 3 5 The size of a dot is defined to be V2 side of a dot Keshidah A procedure applicable to some letters in Persian alphabet which will result in stretching of the letter A stretched letter s length should be between 7 to 11 dots giving the author the right to choose what 1s the most proper length Persian Also known as Farsi The language spoken in Iran Tajikistan and Afkanistan Contains 32 letters 6 CK 4 es B h Q4 a c9 BT 2 ox C CC C e cx Ge ek j ede 96 Oe pt de Be i Most of the Persian letters have up to 4 different shapes e Stand alone e Connected before e Connected after e Connected both before and after Nastalig an Old Persian calligraphy style Spacing The required space we need between not connected letters and also words in Persian Nastaliq It is mainly equal to one dot
74. t have proper curves I define a proper curve as one whose start and end points are positioned in its glyph s path so that it is easy to decide if the glyph is a 6 point or a 4 point stretchable glyph For instance the glyph gt Pe shown in Figure 45 can be easily determined as a 6 point stretchable glyph Curves A B C and D are the four proper curves in this figure Also the glyph shown in Figure 46 can be easily determined as a 4 point stretchable glyph with curves A and B in the figure being the two stretchable curves On the other hand the glyph lt in Figure 47 is one for which it is not easy to decide what type of stretchable glyph it is As a result due to the unsuitable position of the two control points of the curves showed within the larger circles in the outline neither of the two decisions would look perfect after stretching the letters Figure 48 and 49 show the two decisions made The problem with Figure 48 1s that the curves A and C are located higher than where they should be In Figure 49 curves B and D are too close to each other As can be seen the start and end points of a curve in a glyph are proper when their vertical vector is small in comparison to their horizontal vector There 1s no fixed ratio to define this comparison Therefore human interaction 1s needed to define whether the curve is a proper one or not by choosing which one looks more beautiful 73 z D ININ MAY Figure 45 A 6 Point Stret
75. tching in A Hebrew B Arabic Naskh and C Nastaliq 1 3 Problems with Automating Persian Writing and Persian Calligraphy This work was originally focused on enhancing ditroff ffortid 2 to be able to handle Persian Nastaliq calligraphy by using a technique for implementing stretching developed by Abdelouahad Bayar and Khalid Sami 3 and by using a dynamic font 4 to implement the stretching The reason ditroff ffortid was chosen as the main editor for typing Nastaliq is that its modular architecture described in 2 1 1 For example if we need to make changes to the program handling right to left typing no other program should be touched This makes it easier for the developers to add new feature to the editor For this purpose we would need a font containing all the shapes of all glyphs with every possible stretching amount achieved dynamically during printing For this purpose the user manual for a program called NasTroff was prepared to serve as a requirements specification However as the research advanced it became apparent that Nastaliq is much more complicated than other typefaces and that a dynamic font could not be used to achieve the needed stretching This complexity can be categorized into existence of both vertical and horizontal movement of the letters existence of tens of different glyphs for each letter and at last existence of exceptional cases for most of the glyphs Therefore the focus shifted to developing only a font
76. ter None ORSign gt This Grammar was developed to generate strings that encode glyphs in Nastaliq 66 29 Since 1 V and are meaningful characters in PostScript they cannot be used in the naming of glyphs Therefore I have used O instead of and C instead of for grouping The Letter Q means that the current letter is followed by another letter or group of letters Since the or sign e is not a meaningful character in PostScript it can be used to denote or in a glyph name 6 7 Calculating Font Metrics The next step was to calculate each glyph s bounding box and movements This was important because without these information the font could not be used as expected To calculate a glyph s bounding box it is necessary to determine the minimum and maximum 59 values of both the x and y coordinates The horizontal movement was taken from the afm file that was generated by ttf2ptl As discussed in Section 6 3 the vertical information was not given by any conversion tools Therefore all work on the vertical movement of each glyph is postponed until the time that a conversion tool handling both vertical and horizontal information is developed To help me determine bounding boxes I modified the font definition so that it would draw a small circle at the start and end points of each curve and line It drew the first circle in each glyph larger than the othe
77. tters Be Pe ATOLO EY 57 Figure 35 Non Stretched Letter Te Which is Followed by a Letter e Mim Which is Not Connected Following to Any Other Letters eeeeeeeseeeeeeeeen 57 Figure 36 A 4 Point Stretchable Curve Glyph ou cccccccccccceeeeseseeeeeeeeeeseaeeenees 62 Figure 37 A 6 Point Stretchable Curve Glyph eccccccccccccceeeseseeeeeceeeeeeaeeesees 64 Figure 38 Stretching and Component Shifts Letter Be eeeeesssss 67 Figure 39 Stretching and Component Shifts Letter Kaf uueeesesssesssss 68 Figure 40 Stretching and Component Shifts Letter gaf seeeeeeeesssssss 69 Figure 41 Stand Alone Letter lt Kaf Not stretched eseeeesessssss 69 Figure d2 Stretched Letter Il ceret sete iti ol sins reise t Pa Fete ure pre coche 70 Figure 43 Connecting Both Letter Fe of the Size 5 Points sssssse 70 Figure 44 The Connecting Both Letter Fe Stretched Version of Figure 43 by the AMOU OF 4 POI udo nieto a danse essen rM END D A Esa n a ORE 71 Figure 45 A 6 Point Stretchable Letter Pe an Easily Decidable 6 Point btrctchable OIVDDE per de tup onto abbat ede PHA ert oda ern exe but iu dac ERU ede dba daa teb abun uis 74 Figure 46 A 4 Point Stretchable Letter Fe an Easily Decidable 4 Point Siretchable CIDVDDsoo i eto de ote tO MN C MI aal te cd M n eod cam Md te A
78. uSUN UO UDINE MUNI MIND ME UMEN BEN Rensi 90 PA MEMO m 92 2 4 Rules in Persian Nastaliq This section just contains the actual rules in Persian Nastaliq Chapter 4 is about what I will and can implement 95 Roles needed 10 De taken e are OL oosonsetus Qaot Da ebu mE UII M MINIS ONDE E 95 23 E e e S s BASIC USO Cds Sans E qM UNUM 99 AE Dover ok Gil CUC WIN ueni UPS rU OON I HEISE DI PT INE OR tete 99 S NN NOE ITO IE SRL RTT 99 ELO IE M 99 t CST IC CIS Oea EA E E E E 100 OVE AC ETES E EEEE 100 S MEE esi NR 100 Adl A tomauc GANG EQUIV m 100 z SAP Mowo libi NE 102 TREE SIN TTTRTE 102 FIGURE 1 THE FLOW OF DATA FROM KEYBOARD TO OUTPUT RESULT 88 FIGURE 2 HIGH LEVEL VIEW OF NAS AND TROFF PROCESSORS 89 FIGURE 3 SAMPLE PERSIAN NASTALIQ EDITING WITH TROFF 90 86 1 Introduction 1 1 Product Overview NAS is a Troff postprocessor for handling the Persian calligraphy called Nastaliq NAS lets Troff documents contain Nastalig as well as automatically handling the stretching problem while both right and left justification In Chapter 2 the required rules for Persian Nastaliq as well as the rules handling the stretching problem are described Chapter 3 describes some basic use cases about stretching Chapter 4 provides the algorithms and procedure NAS uses to stretch letter This software is meant to run on an Apples Mac OS X

(TITLE OF THE THESIS)* - UWSpace

Contents

Download Pdf Manuals

Related Search

Related Contents