PhilT_eX

Chances are, if you’re an aca­d­e­mic in the 21^st cen­tury, you read a lot of PDFs, whether they are arti­cles down­loaded from jour­nals’ web­sites, works in progress from col­leagues, or a stu­dent papers. To avoid the costs of print­ing, and for your own con­ve­nience, you might be inter­ested in read­ing them on a portable device, such as e-book reader like the Sony Reader Dig­i­tal Book, Barnes & Noble Nook or Ama­zon Kin­dle. Although larger mod­els exist, the rea­son­ably priced ones have screens the size of small paper­backs, roughly 9 × 12cm. (Of course, you might even want to use a smart phone with an even smaller screen!) This can pose a prob­lem when it comes to PDFs designed to be printed on let­ter (8.5 × 11in) or even A4 sized paper. The device will typ­i­cally shrink them to fit, result­ing in an unread­ably small font size.

But there are a num­ber of things you can try to make those PDFs more read­able. Here I focus my com­ments on free meth­ods, rather than on those that rely on expen­sive com­mer­cial tools like Adobe Acro­bat Pro­fes­sional (which, to be hon­est, doesn’t work that well for this any­way). Here’s a run down of my expe­ri­ences with a vari­ety of methods:

1. Work­ing with the source doc­u­ment, if available
2. Sim­ple white­space cropping
3. Spe­cial crop­ping tools like SoPDF and BRISS
4. “Reflow­ing” and/or con­vert­ing to another format

---

Work­ing with the source doc­u­ment, if available

This is by far the best method. The catch is obvi­ous: you actu­ally have to have access to the source doc­u­ment from which the PDF was cre­ated. This is great for your own works, and per­haps those of stu­dents or gen­er­ous col­leagues, but often not even a pos­si­bil­ity. For a tra­di­tional Word­Proces­sor file, it usu­ally requires chang­ing the out­put “paper” size to the size of your portable, and min­i­miz­ing the mar­gins. Sony has put together excel­lent and detailed instruc­tions for doing this with OpenOf­fice and/or Microsoft Word, which would mostly carry over to other portables.

Read­ers of this blog may be inter­ested in doing it with LaTeX source files. I’ve had luck adding the fol­low­ing (or a slight mod­i­fi­ca­tion) to the preamble:

\usepackage[papersize={90mm,120mm},margin=2mm]{geometry}
\usepackage[kerning=true]{microtype}
\usepackage[T1]{fontenc}
\usepackage[charter]{mathdesign}
\usepackage[normalmargins]{savetrees}
\sloppy
\pagestyle{empty}

The geom­e­try pack­age sets the page size and very min­i­mal mar­gins. You’ll want to change the first option to match your device’s screen, and adjust the sec­ond accord­ing to pref­er­ence. The microtype pack­age improves typog­ra­phy in cramped quar­ters, and allows cer­tain punc­tu­a­tion to extend slightly into the mar­gins in a way that is vis­i­bly unno­tice­able. The save­trees option adjusts the size and spac­ing of titles and lists, and gen­er­ally squeezes more onto a page. (The nor­mal­mar­gins option is needed to avoid a clash with our geom­e­try set­tings.) The \sloppy com­mand makes LaTeX less fussy about avoid­ing under­full “bad boxes”; with­out this you’ll get text run­ning off the right side of the screen. The \pagestyle com­mand sup­presses head­ers and foot­ers and page num­bers. (Not that there’s room for them anyway.)

I like to use T1 font encod­ing with the Bit­stream Char­ter font pro­vided by the math­de­sign pack­age. I find this font to be very leg­i­ble when read from a screen, but still pro­fes­sional look­ing. This is of course a mat­ter of taste. Sub­sti­tute your favorite font instead. Per­son­ally, I think the default Com­puter Mod­ern LaTeX fonts are bet­ter suited for paper than screen read­ing. If you choose to use XeLa­TeX for greater font selec­tion, go ahead, but you’ll need to remove the microtype pack­age, which only works with reg­u­lar latex/pdflatex.

The result should be some­thing quite nice and com­fort­able to read on a portable device. If you visit dis­cus­sion com­mu­ni­ties ded­i­cated to e-book read­ers like Mobil­eRead or Tel­eRead, you’ll read a lot of bash­ing of the PDF for­mat. In my opin­ion, PDF is both the best and the worst doc­u­ment for­mat. It’s the best when the PDF was designed for the medium it’s viewed on. It’s the worst when it was not.

---

Sim­ple white­space cropping

One of the sim­plest changes you can make to a PDF, with­out chang­ing any of the con­tent or for­mat­ting, is to remove the mar­gins so that the con­tent of the pages takes up the whole screen. There is a lot of soft­ware, both free and com­mer­cial, that can be used to crop a PDF, includ­ing pdfla­tex itself (though it’s not the most effi­cient tool for the job). Unfor­tu­nately, it’s not always as straight­for­ward as it may seem. The so-called “crop” fea­ture of Adobe Acro­bat Pro­fes­sional, for exam­ple, doesn’t actu­ally crop PDFs, but instead inserts an instruc­tion telling the view­ing soft­ware to ignore the cropped areas, which often isn’t respected by portable devices.

Nonethe­less, you prob­a­bly already have free/open source soft­ware installed that can do the trick. If you have LaTeX installed, you almost cer­tainly have ghost­script installed. If you own a dig­i­tal e-book reader, there’s a good chance you have cal­i­bre, the free/open source ebook man­age­ment soft­ware installed. (And if not, you should, even if you don’t like it as a library man­ager.) The two of them together can be used to ana­lyze and auto-crop a PDF with a sin­gle com­mand, or even batch-crop an entire folder’s worth.

On linux/Unix (includ­ing, I think, Mac OS X, though I haven’t tested this), from the com­mand line, you can type in:

for i in *.pdf ; do gs -dSAFER -dNOPAUSE -dBATCH -sDEVICE=bbox "$i" 2> bounding ; pdfmanipulate crop -o "${i%.pdf}-cropped.pdf" -b bounding "$i"; done

This will cre­ate a cropped ver­sion of every PDF in the active directory.

On Win­dows, you can do the same by cre­at­ing a batch file. Using notepad or other text edi­tor, cre­ate a file named pdfbatchcrop.bat (or any­thing else you’ll remem­ber end­ing with .bat), and copy and paste the fol­low­ing into it:

for %%I in (*.pdf) do (
"C:\Program Files\gs\gs8.71\bin\gswin32c.exe" -dSAFER -dNOPAUSE -dBATCH -sDEVICE#bbox "%%I" 2> bounding
"C:\Program Files\Calibre2\pdfmanipulate.exe" crop -o "%%~nI-cropped.pdf" -b bounding "%%I"
)

(Dou­ble check that the paths I have given to your cal­i­bre and ghost­script exe­cuta­bles are right. They may not be if you’re using an older ver­sion of ghost­script, or a 64-bit ver­sion of Win­dows.) Save it in the same direc­tory as the PDFs you want to crop, and then dou­ble click on the .bat file in “My Com­puter” or “Win­dows Explorer”, and it should crop away.

If a sim­ple crop isn’t “cut­ting it” for you, whether it is because the header and footer are not removed, or the PDFs have mul­ti­ple columns which need sep­a­rat­ing, read on for other suggestions.

---

Spe­cial crop­ping tools like SoPDF and BRISS

If you need a “smarter” crop, or if you’re one of those peo­ple to whom com­mand line pro­grams are anath­ema, there are some free tools out there that are espe­cially help­ful in crop­ping PDFs, or even divid­ing PDF pages into e-reader-screen-sized “chunks”. Here I focus on two of my favorites.

SoPDF

SoPDF is a com­mand line pro­gram, writ­ten in C/C++, avail­able for both Win­dows and Linux. (The source code is avail­able, if any­one wants to try com­pil­ing it for Mac.) For those who hate the com­mand line, there are also mul­ti­ple GUI wrap­pers for it, one of which, avail­able here, in full dis­clo­sure, was writ­ten by yours truly back when I still used Win­dows. (Another, in some ways bet­ter, cross-platform GUI can be found here.)

It not only auto-crops a PDF, but also (at least on default set­tings), divides each page in half and rotates it 90 degrees. This is meant to allow you to read com­fort­ably a PDF, half a page at a time, by hold­ing your reader side­ways to take advan­tage of the high height/width pro­por­tion of these devices.

Despite the GUIs avail­able, on linux, I often find myself using the com­mand line ver­sion for batch pro­cess­ing. E.g., to do a direc­tory of PDFs at once:

for file in *.pdf ; do sopdf -i "$file" ; done

SoPDF is light­ning fast. Com­pared to the ghostscript/calibre method men­tioned above, you won’t believe how fast this thing works. With its quick and effec­tive method, it’s by far the most-often used tool in my tool­box for mak­ing PDFs read­able on my Sony Reader.

The down­sides are that the orig­i­nal devel­oper has stopped work­ing on it, so it’s stag­nated a bit, and doesn’t work if you need to cus­tomize the crop amount (for mul­ti­ple columns, or to remove large head­ers and foot­ers). In these cases, I turn to …

BRISS

A new and very promis­ing open source tool on the scene is BRISS. The devel­oper likes to call it “a lit­tle snip­ping tool to make the most of your six inch.” (A ref­er­ence to the six inch screen of these portable devices.) Yes, I apol­o­gize for repeat­ing that.

This is a cross-platform Java tool with a nice GUI. It shows you a sort of fuzzy blend of all the pages at once, so that you can man­u­ally select crop regions for divid­ing a page into one, two or more chunks. It also does intel­li­gent han­dling of doc­u­ments with dif­fer­ent mar­gins on odd and even pages, so that you can treat them differently.

Here’s a screenshot.

In fact, I can imag­ine this tool being use­ful in just about any con­text in which you might want to crop a PDF, not just for the pur­pose of read­ing PDFs on porta­bles. One con­text that comes to mind is the pos­si­ble need for extract­ing a sin­gle image or dia­gram from a PDF with­out con­vert­ing to a ras­ter­ized format.

BRISS is a new project and under rapid devel­op­ment. I expect new fea­tures, such as page rota­tion, in the future.

The are some older tools with sim­i­lar func­tions. The big down­side with these is that they ras­ter­ize the PDF, i.e., con­vert it to a series of pix­elized images, before doing other pro­cess­ing. This has the down­side of increas­ing file size quite a lot, and also makes it impos­si­ble to use search or dic­tio­nary fea­tures of your reader, if it has them. Nev­er­the­less, espe­cially if your PDF starts off as a series of images (such as a scan), these might be worth look­ing at:

• PDF­Read: PDF read­ing solu­tion avail­able for Mac and Windows.
• Paper­Crop: Spe­cial­ized tool for crop­ping and divid­ing multi-column PDFs (Win­dows only?)
• PDFLRF: Very nice page divi­sion and even recom­bi­na­tion algo­rithm, but out­puts a rather out­dated for­mat, and is no longer being updated

---

“Reflow­ing” and/or con­vert­ing to another format

Some portable read­ers adver­tise a fea­ture to “reflow” a PDF, or take the words and sen­tences and rearrange them to match their dis­play, and your font size pref­er­ence. In my expe­ri­ence, at least with my (admit­tedly a few years old) Sony Reader PRS-505, this does not work well at all. Often the result is an unread­able mess, and parts of the doc­u­ment (espe­cially vec­tor graph­ics) are sim­ply lost.

This isn’t all that sur­pris­ing. The PDF for­mat wasn’t designed to allow the end user to cus­tomize the look of one. Instead, it was designed pre­cisely to pre­serve a uni­form look across the board. In other words, it was designed purely as an out­put for­mat. A PDF basi­cally stores the exact loca­tion of each char­ac­ter or image and noth­ing else. It doesn’t even store infor­ma­tion about what char­ac­ters go together to make up a sin­gle word, much less where para­graphs begin and end, what con­sti­tutes a header or foot­note as opposed to the body text. It requires a fairly high level of arti­fi­cial intel­li­gence for a com­puter to recon­struct such infor­ma­tion, which of course is nec­es­sary to con­vert a PDF to a reflow­able format.

It’s prob­a­bly not worth even try­ing to con­vert a PDF to another for­mat if it con­tains com­plex for­mat­ting such as dia­grams, tables, log­i­cal and math­e­mat­i­cal for­mu­las or char­ac­ters from non-Latin alpha­bets. But if it’s more or less straight Latin alpha­bet text arranged in nor­mal para­graphs, it might be worth giv­ing it a shot, espe­cially if you can live with less-than per­fect results.

Again, I’m lim­it­ing my dis­cus­sion to free tools. I believe all of the fol­low­ing rely on the open source pop­pler PDF libraries for their basic tex­tual analy­sis, and hence, yield fairly sim­i­lar results.

Abi­Word

Abi­Word is actu­ally an open source cross-platform word proces­sor. How­ever, espe­cially if you install the addi­tional import and export plu­g­ins, it is very handy for con­vert­ing between for­mats. (It can even be used from the com­mand line to con­vert between any for­mats it knows, which by itself makes it an invalu­able tool. It even has a LaTeX export plu­gin, so you can use it to con­vert, e.g., MS Word docs and sim­i­lar to LaTeX, but that’s a topic for another occasion …)

Abi­Word will open a PDF and attempt to con­vert it into a Word Proces­sor file that can be edited as nor­mal. The results are usu­ally far from per­fect, but this is still very handy in a vari­ety of cir­cum­stances. You can then save the result as RTF or HTML or ODT. Some read­ers sup­port such for­mats directly, but if yours doesn’t, these for­mats are much eas­ier to con­vert to native e-Reader for­mats like ePub or mobi.

(A newer plu­gin for OpenOf­fice that does more or less the same is also avail­able, though I have lit­tle expe­ri­ence myself.)

Cal­i­bre

Men­tioned above, cal­i­bre is an amaz­ingly rich e-book soft­ware suite that can be used for library man­age­ment, doc­u­ment view­ing and con­ver­sion between many dif­fer­ent e-book for­mats. It comes with a fairly user-friendly (if some­what slow and uncus­tomiz­able) graph­i­cal user inter­face, but most of its func­tions can also be called from a command-line shell, which is, again, great for scripts.

Cal­i­bre recently added the option of con­vert­ing directly from PDF to for­mats such as ePub (used by most e-book read­ers, not to men­tion the iPad’s iBooks appli­ca­tion) or mobi (one of the few for­mats sup­ported by the Kin­dle). It allows you to use Reg­u­lar Expres­sions for header and footer detec­tion and removal. Again, the results are usu­ally far from per­fect, but for a quick con­ver­sion of a rel­a­tively sim­ple doc­u­ment, this is a very sim­ple and ele­gant solution.

pdfre­flow

Another rel­a­tively new project, pdfre­flow is a com­mand line pro­gram (writ­ten in Java and avail­able for Win­dows, Mac and Linux) that can be used to “reflow” the HTML out­put of poppler’s pdfto­html con­ver­sion rou­tine. (Really, it con­verts an XML file into a more tra­di­tional HTML file than pdfto­html gen­er­ates by itself.) There are more detailed instruc­tions for it given here.

While many peo­ple dis­like command-line only pro­grams such as this, it is good for script­ing, and in gen­eral, I’ve found that its para­graph and for­mat­ting detec­tion algo­rithms seem to be more reli­able than calibre’s. Its recog­ni­tion of things like foot­notes could be improved, but per­haps will in future releases.

Since it gen­er­ates HTML, you may need to use other soft­ware for con­vert­ing the HTML into some­thing your e-book reader can han­dle, but HTML is an ideal for­mat for con­vert­ing to oth­ers. Con­vert­ing to ebook for­mats can be done fairly eas­ily with cal­i­bre or a WYSIWYG ePub edi­tor such as Sigil.

A final thing to con­sider is com­bin­ing tools. For exam­ple, using BRISS to crop away head­ers and foot­ers, and then con­vert­ing the results with Abi­Word, cal­i­bre or pdfre­flow might yield bet­ter results than one of these tools alone.