Problem saving to eps with usetex=True

I've been trying to track down some strange behavior I was getting,
and I think narrowed it down to some code that I'll paste below.

I'm trying to write to .eps files, and when I have usetex=True,
something is screwed up with the padding on the left, and eventually
the whole image is just white.

If I run this script, the 'testA-*.eps' look good, but 'testB-*' does
not. The same problem happens even if I remove the ticklabels.

Any tips would be appreciated.
thanks,
Ken

···

------------------------------------------------
import matplotlib.pyplot as plt
import numpy as np
from matplotlib import rc

fig = plt.figure()
ax = fig.add_axes([0,0,1,1],frameon=False)

X = np.tile(np.arange(500),(10,1)) # (10,500) shape

ax.imshow(X,interpolation='nearest',aspect='auto')

def go(name):

    for d in (1,2,3,4):

        w = d*5
        h = d

        fig.set_size_inches(w,h)
        fig.savefig("%s-%d.eps" % (name,d))

rc('text', usetex=False)
go("testA")

rc('text', usetex=True)
go("testB")

ps backend, when usetex=True, uses latex with psfrag package to
generate the output (with some extra steps).
It seems that the bounding box information is not correctly recovered
during this process.
I first thought that it would be quite difficult to get this correct,
however the attached (relatively simple) patch seems to work fine.

Ken, can you try the patch and see if it works?

-JJ

ps.diff (1.95 KB)

···

On Thu, Apr 23, 2009 at 2:25 PM, Ken Schutte <kts.lists@...149...> wrote:

I've been trying to track down some strange behavior I was getting,
and I think narrowed it down to some code that I'll paste below.

I'm trying to write to .eps files, and when I have usetex=True,
something is screwed up with the padding on the left, and eventually
the whole image is just white.

If I run this script, the 'testA-*.eps' look good, but 'testB-*' does
not. The same problem happens even if I remove the ticklabels.

Any tips would be appreciated.
thanks,
Ken

------------------------------------------------
import matplotlib.pyplot as plt
import numpy as np
from matplotlib import rc

fig = plt.figure()
ax = fig.add_axes([0,0,1,1],frameon=False)

X = np.tile(np.arange(500),(10,1)) # (10,500) shape

ax.imshow(X,interpolation='nearest',aspect='auto')

def go(name):

for d in (1,2,3,4):

   w = d\*5
   h = d

   fig\.set\_size\_inches\(w,h\)
   fig\.savefig\(&quot;%s\-%d\.eps&quot; % \(name,d\)\)

rc('text', usetex=False)
go("testA")

rc('text', usetex=True)
go("testB")

------------------------------------------------------------------------------
Crystal Reports &#45; New Free Runtime and 30 Day Trial
Check out the new simplified licensign option that enables unlimited
royalty&#45;free distribution of the report engine for externally facing
server and web deployment.
http://p.sf.net/sfu/businessobjects
_______________________________________________
Matplotlib-devel mailing list
Matplotlib-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

Yeah, that seems to work! thanks a lot,
Ken

···

On Fri, Apr 24, 2009 at 5:21 PM, Jae-Joon Lee <lee.j.joon@...149...> wrote:

ps backend, when usetex=True, uses latex with psfrag package to
generate the output (with some extra steps).
It seems that the bounding box information is not correctly recovered
during this process.
I first thought that it would be quite difficult to get this correct,
however the attached (relatively simple) patch seems to work fine.

Ken, can you try the patch and see if it works?

-JJ

On Thu, Apr 23, 2009 at 2:25 PM, Ken Schutte <kts.lists@...149...> wrote:

I've been trying to track down some strange behavior I was getting,
and I think narrowed it down to some code that I'll paste below.

I'm trying to write to .eps files, and when I have usetex=True,
something is screwed up with the padding on the left, and eventually
the whole image is just white.

If I run this script, the 'testA-*.eps' look good, but 'testB-*' does
not. The same problem happens even if I remove the ticklabels.

Any tips would be appreciated.
thanks,
Ken

------------------------------------------------
import matplotlib.pyplot as plt
import numpy as np
from matplotlib import rc

fig = plt.figure()
ax = fig.add_axes([0,0,1,1],frameon=False)

X = np.tile(np.arange(500),(10,1)) # (10,500) shape

ax.imshow(X,interpolation='nearest',aspect='auto')

def go(name):

for d in (1,2,3,4):

   w = d\*5
   h = d

   fig\.set\_size\_inches\(w,h\)
   fig\.savefig\(&quot;%s\-%d\.eps&quot; % \(name,d\)\)

rc('text', usetex=False)
go("testA")

rc('text', usetex=True)
go("testB")

------------------------------------------------------------------------------
Crystal Reports &#45; New Free Runtime and 30 Day Trial
Check out the new simplified licensign option that enables unlimited
royalty&#45;free distribution of the report engine for externally facing
server and web deployment.
http://p.sf.net/sfu/businessobjects
_______________________________________________
Matplotlib-devel mailing list
Matplotlib-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

This patch is now committed to the trunk (r7068).

-JJ

···

On Sat, Apr 25, 2009 at 8:29 AM, Ken Schutte <kts.lists@...149...> wrote:

Yeah, that seems to work! thanks a lot,
Ken

On Fri, Apr 24, 2009 at 5:21 PM, Jae-Joon Lee <lee.j.joon@...149...> wrote:

ps backend, when usetex=True, uses latex with psfrag package to
generate the output (with some extra steps).
It seems that the bounding box information is not correctly recovered
during this process.
I first thought that it would be quite difficult to get this correct,
however the attached (relatively simple) patch seems to work fine.

Ken, can you try the patch and see if it works?

-JJ

On Thu, Apr 23, 2009 at 2:25 PM, Ken Schutte <kts.lists@...149...> wrote:

I've been trying to track down some strange behavior I was getting,
and I think narrowed it down to some code that I'll paste below.

I'm trying to write to .eps files, and when I have usetex=True,
something is screwed up with the padding on the left, and eventually
the whole image is just white.

If I run this script, the 'testA-*.eps' look good, but 'testB-*' does
not. The same problem happens even if I remove the ticklabels.

Any tips would be appreciated.
thanks,
Ken

------------------------------------------------
import matplotlib.pyplot as plt
import numpy as np
from matplotlib import rc

fig = plt.figure()
ax = fig.add_axes([0,0,1,1],frameon=False)

X = np.tile(np.arange(500),(10,1)) # (10,500) shape

ax.imshow(X,interpolation='nearest',aspect='auto')

def go(name):

for d in (1,2,3,4):

   w = d\*5
   h = d

   fig\.set\_size\_inches\(w,h\)
   fig\.savefig\(&quot;%s\-%d\.eps&quot; % \(name,d\)\)

rc('text', usetex=False)
go("testA")

rc('text', usetex=True)
go("testB")

------------------------------------------------------------------------------
Crystal Reports &#45; New Free Runtime and 30 Day Trial
Check out the new simplified licensign option that enables unlimited
royalty&#45;free distribution of the report engine for externally facing
server and web deployment.
http://p.sf.net/sfu/businessobjects
_______________________________________________
Matplotlib-devel mailing list
Matplotlib-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

Hi Jae-Joon,

This patch is now committed to the trunk (r7068).

I think these changes have had unintended consequences. I attached a test file I used to inspect the results when I wrote the original code that you recently tried to improve. There are lots of combinations of large and small figures, with usetex and without, and with no distiller, gs distiller, of xpdf (pdftops) distiller. You have to visually inspect all of the ps/eps files that are generated. At one time, all of these files looked acceptable, but that was a couple years ago and I don’t remember what version of ghostscript I was using at the time (I’m sure there is a record of it somewhere on this mailing list). WIth r7067, most of the files look fine, a few of the large files (large paper sizes) have problems. With r7068, quite a few examples have problems.

Darren

···

On Tue, Apr 28, 2009 at 3:49 PM, Jae-Joon Lee <lee.j.joon@…149…> wrote:

On Sat, Apr 25, 2009 at 8:29 AM, Ken Schutte <kts.lists@…149…> wrote:

Yeah, that seems to work! thanks a lot,

Ken

On Fri, Apr 24, 2009 at 5:21 PM, Jae-Joon Lee <lee.j.joon@…149…> wrote:

ps backend, when usetex=True, uses latex with psfrag package to

generate the output (with some extra steps).

It seems that the bounding box information is not correctly recovered

during this process.

I first thought that it would be quite difficult to get this correct,

however the attached (relatively simple) patch seems to work fine.

Ken, can you try the patch and see if it works?

-JJ

On Thu, Apr 23, 2009 at 2:25 PM, Ken Schutte <kts.lists@…149…> wrote:

I’ve been trying to track down some strange behavior I was getting,

and I think narrowed it down to some code that I’ll paste below.

I’m trying to write to .eps files, and when I have usetex=True,

something is screwed up with the padding on the left, and eventually

the whole image is just white.

If I run this script, the ‘testA-.eps’ look good, but 'testB-’ does

not. The same problem happens even if I remove the ticklabels.

Any tips would be appreciated.

thanks,

Ken


import matplotlib.pyplot as plt

import numpy as np

from matplotlib import rc

fig = plt.figure()

ax = fig.add_axes([0,0,1,1],frameon=False)

X = np.tile(np.arange(500),(10,1)) # (10,500) shape

ax.imshow(X,interpolation=‘nearest’,aspect=‘auto’)

def go(name):

for d in (1,2,3,4):

   w = d*5
   h = d
   fig.set_size_inches(w,h)
   fig.savefig("%s-%d.eps" % (name,d))

rc(‘text’, usetex=False)

go(“testA”)

rc(‘text’, usetex=True)

go(“testB”)


Crystal Reports - New Free Runtime and 30 Day Trial

Check out the new simplified licensign option that enables unlimited

royalty-free distribution of the report engine for externally facing

server and web deployment.

http://p.sf.net/sfu/businessobjects


Matplotlib-devel mailing list

Matplotlib-devel@lists.sourceforge.net

https://lists.sourceforge.net/lists/listinfo/matplotlib-devel


Register Now & Save for Velocity, the Web Performance & Operations

Conference from O’Reilly Media. Velocity features a full day of

expert-led, hands-on workshops and two days of sessions from industry

leaders in dedicated Performance & Operations tracks. Use code vel09scf

and Save an extra 15% before 5/3. http://p.sf.net/sfu/velocityconf


Matplotlib-devel mailing list

Matplotlib-devel@lists.sourceforge.net

https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

The test file is attached this time.

usetex_ps_test.py (6.36 KB)

···

On Wed, Apr 29, 2009 at 3:55 PM, Darren Dale <dsdale24@…149…> wrote:

Hi Jae-Joon,

On Tue, Apr 28, 2009 at 3:49 PM, Jae-Joon Lee <lee.j.joon@…149…> wrote:

This patch is now committed to the trunk (r7068).

I think these changes have had unintended consequences. I attached a test file I used to inspect the results when I wrote the original code that you recently tried to improve. There are lots of combinations of large and small figures, with usetex and without, and with no distiller, gs distiller, of xpdf (pdftops) distiller. You have to visually inspect all of the ps/eps files that are generated. At one time, all of these files looked acceptable, but that was a couple years ago and I don’t remember what version of ghostscript I was using at the time (I’m sure there is a record of it somewhere on this mailing list). WIth r7067, most of the files look fine, a few of the large files (large paper sizes) have problems. With r7068, quite a few examples have problems.

Darren

On Sat, Apr 25, 2009 at 8:29 AM, Ken Schutte <kts.lists@…149…> wrote:

Yeah, that seems to work! thanks a lot,

Ken

On Fri, Apr 24, 2009 at 5:21 PM, Jae-Joon Lee <lee.j.joon@…149…> wrote:

ps backend, when usetex=True, uses latex with psfrag package to

generate the output (with some extra steps).

It seems that the bounding box information is not correctly recovered

during this process.

I first thought that it would be quite difficult to get this correct,

however the attached (relatively simple) patch seems to work fine.

Ken, can you try the patch and see if it works?

-JJ

On Thu, Apr 23, 2009 at 2:25 PM, Ken Schutte <kts.lists@…149…> wrote:

I’ve been trying to track down some strange behavior I was getting,

and I think narrowed it down to some code that I’ll paste below.

I’m trying to write to .eps files, and when I have usetex=True,

something is screwed up with the padding on the left, and eventually

the whole image is just white.

If I run this script, the ‘testA-.eps’ look good, but 'testB-’ does

not. The same problem happens even if I remove the ticklabels.

Any tips would be appreciated.

thanks,

Ken


import matplotlib.pyplot as plt

import numpy as np

from matplotlib import rc

fig = plt.figure()

ax = fig.add_axes([0,0,1,1],frameon=False)

X = np.tile(np.arange(500),(10,1)) # (10,500) shape

ax.imshow(X,interpolation=‘nearest’,aspect=‘auto’)

def go(name):

for d in (1,2,3,4):

   w = d*5
   h = d
   fig.set_size_inches(w,h)
   fig.savefig("%s-%d.eps" % (name,d))

rc(‘text’, usetex=False)

go(“testA”)

rc(‘text’, usetex=True)

go(“testB”)


Crystal Reports - New Free Runtime and 30 Day Trial

Check out the new simplified licensign option that enables unlimited

royalty-free distribution of the report engine for externally facing

server and web deployment.

http://p.sf.net/sfu/businessobjects


Matplotlib-devel mailing list

Matplotlib-devel@lists.sourceforge.net

https://lists.sourceforge.net/lists/listinfo/matplotlib-devel


Register Now & Save for Velocity, the Web Performance & Operations

Conference from O’Reilly Media. Velocity features a full day of

expert-led, hands-on workshops and two days of sessions from industry

leaders in dedicated Performance & Operations tracks. Use code vel09scf

and Save an extra 15% before 5/3. http://p.sf.net/sfu/velocityconf


Matplotlib-devel mailing list

Matplotlib-devel@lists.sourceforge.net

https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

Yes, I'm sorry, I also spoke too soon. I have found problems with
this patch - mainly things like it chopping off xlabel, titles, etc,
for some examples.

I have some major deadlines using this stuff right now, so I don't
have time to experiment. But, eventually, I will try to take a look
at what is happening.

Ken

···

On Wed, Apr 29, 2009 at 3:56 PM, Darren Dale <dsdale24@...149...> wrote:

The test file is attached this time.

On Wed, Apr 29, 2009 at 3:55 PM, Darren Dale <dsdale24@...149...> wrote:

Hi Jae-Joon,

On Tue, Apr 28, 2009 at 3:49 PM, Jae-Joon Lee <lee.j.joon@...149...> >> wrote:

This patch is now committed to the trunk (r7068).

I think these changes have had unintended consequences. I attached a test
file I used to inspect the results when I wrote the original code that you
recently tried to improve. There are lots of combinations of large and small
figures, with usetex and without, and with no distiller, gs distiller, of
xpdf (pdftops) distiller. You have to visually inspect all of the ps/eps
files that are generated. At one time, all of these files looked acceptable,
but that was a couple years ago and I don't remember what version of
ghostscript I was using at the time (I'm sure there is a record of it
somewhere on this mailing list). WIth r7067, most of the files look fine, a
few of the large files (large paper sizes) have problems. With r7068, quite
a few examples have problems.

Darren

On Sat, Apr 25, 2009 at 8:29 AM, Ken Schutte <kts.lists@...149...> wrote:
> Yeah, that seems to work! thanks a lot,
> Ken
>
> On Fri, Apr 24, 2009 at 5:21 PM, Jae-Joon Lee <lee.j.joon@...149...> >>> > wrote:
>> ps backend, when usetex=True, uses latex with psfrag package to
>> generate the output (with some extra steps).
>> It seems that the bounding box information is not correctly recovered
>> during this process.
>> I first thought that it would be quite difficult to get this correct,
>> however the attached (relatively simple) patch seems to work fine.
>>
>> Ken, can you try the patch and see if it works?
>>
>> -JJ
>>
>>
>>
>>
>> On Thu, Apr 23, 2009 at 2:25 PM, Ken Schutte <kts.lists@...149...> >>> >> wrote:
>>> I've been trying to track down some strange behavior I was getting,
>>> and I think narrowed it down to some code that I'll paste below.
>>>
>>> I'm trying to write to .eps files, and when I have usetex=True,
>>> something is screwed up with the padding on the left, and eventually
>>> the whole image is just white.
>>>
>>> If I run this script, the 'testA-*.eps' look good, but 'testB-*' does
>>> not. The same problem happens even if I remove the ticklabels.
>>>
>>> Any tips would be appreciated.
>>> thanks,
>>> Ken
>>>
>>>
>>>
>>> ------------------------------------------------
>>> import matplotlib.pyplot as plt
>>> import numpy as np
>>> from matplotlib import rc
>>>
>>> fig = plt.figure()
>>> ax = fig.add_axes([0,0,1,1],frameon=False)
>>>
>>> X = np.tile(np.arange(500),(10,1)) # (10,500) shape
>>>
>>> ax.imshow(X,interpolation='nearest',aspect='auto')
>>>
>>> def go(name):
>>>
>>> for d in (1,2,3,4):
>>>
>>> w = d*5
>>> h = d
>>>
>>> fig.set_size_inches(w,h)
>>> fig.savefig("%s-%d.eps" % (name,d))
>>>
>>> rc('text', usetex=False)
>>> go("testA")
>>>
>>> rc('text', usetex=True)
>>> go("testB")
>>>
>>>
>>> ------------------------------------------------------------------------------
>>> Crystal Reports &#45; New Free Runtime and 30 Day Trial
>>> Check out the new simplified licensign option that enables unlimited
>>> royalty&#45;free distribution of the report engine for externally
>>> facing
>>> server and web deployment.
>>> http://p.sf.net/sfu/businessobjects
>>> _______________________________________________
>>> Matplotlib-devel mailing list
>>> Matplotlib-devel@lists.sourceforge.net
>>> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
>>>
>>
>

------------------------------------------------------------------------------

Register Now & Save for Velocity, the Web Performance & Operations
Conference from O'Reilly Media. Velocity features a full day of
expert-led, hands-on workshops and two days of sessions from industry
leaders in dedicated Performance & Operations tracks. Use code vel09scf
and Save an extra 15% before 5/3. http://p.sf.net/sfu/velocityconf
_______________________________________________
Matplotlib-devel mailing list
Matplotlib-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

Thanks Darren,

I believe that my patch only affects eps output with usetex=True,
i.e., only those "tex_*.eps" files are affected.

I found that my previous patch didn't treated the orientation of the
paper correctly.
So the bounding box of all the landscape eps outputs are incorrect.
This can be easily fixed.

Index: lib/matplotlib/backends/backend_ps.py

···

===================================================================
--- lib/matplotlib/backends/backend_ps.py (revision 7071)
+++ lib/matplotlib/backends/backend_ps.py (working copy)
@@ -1100,8 +1100,11 @@
         # set the paper size to the figure size if isEPSF. The
         # resulting ps file has the given size with correct bounding
         # box so that there is no need to call 'pstoeps'
             paperWidth, paperHeight = self.figure.get_size_inches()
+ if isLandscape:
+ paperWidth, paperHeight = paperHeight, paperWidth
         else:
             temp_papertype = _get_papertype(width, height)
             if papertype=='auto':

Anyhow, after the landscape fix, all eps output looks fine to me.
Can you check this?

Regards,

-JJ

On Wed, Apr 29, 2009 at 3:55 PM, Darren Dale <dsdale24@...149...> wrote:

Hi Jae-Joon,

On Tue, Apr 28, 2009 at 3:49 PM, Jae-Joon Lee <lee.j.joon@...149...> wrote:

This patch is now committed to the trunk (r7068).

I think these changes have had unintended consequences. I attached a test
file I used to inspect the results when I wrote the original code that you
recently tried to improve. There are lots of combinations of large and small
figures, with usetex and without, and with no distiller, gs distiller, of
xpdf (pdftops) distiller. You have to visually inspect all of the ps/eps
files that are generated. At one time, all of these files looked acceptable,
but that was a couple years ago and I don't remember what version of
ghostscript I was using at the time (I'm sure there is a record of it
somewhere on this mailing list). WIth r7067, most of the files look fine, a
few of the large files (large paper sizes) have problems. With r7068, quite
a few examples have problems.

Darren

On Sat, Apr 25, 2009 at 8:29 AM, Ken Schutte <kts.lists@...149...> wrote:
> Yeah, that seems to work! thanks a lot,
> Ken
>
> On Fri, Apr 24, 2009 at 5:21 PM, Jae-Joon Lee <lee.j.joon@...149...> >> > wrote:
>> ps backend, when usetex=True, uses latex with psfrag package to
>> generate the output (with some extra steps).
>> It seems that the bounding box information is not correctly recovered
>> during this process.
>> I first thought that it would be quite difficult to get this correct,
>> however the attached (relatively simple) patch seems to work fine.
>>
>> Ken, can you try the patch and see if it works?
>>
>> -JJ
>>
>>
>>
>>
>> On Thu, Apr 23, 2009 at 2:25 PM, Ken Schutte <kts.lists@...149...> >> >> wrote:
>>> I've been trying to track down some strange behavior I was getting,
>>> and I think narrowed it down to some code that I'll paste below.
>>>
>>> I'm trying to write to .eps files, and when I have usetex=True,
>>> something is screwed up with the padding on the left, and eventually
>>> the whole image is just white.
>>>
>>> If I run this script, the 'testA-*.eps' look good, but 'testB-*' does
>>> not. The same problem happens even if I remove the ticklabels.
>>>
>>> Any tips would be appreciated.
>>> thanks,
>>> Ken
>>>
>>>
>>>
>>> ------------------------------------------------
>>> import matplotlib.pyplot as plt
>>> import numpy as np
>>> from matplotlib import rc
>>>
>>> fig = plt.figure()
>>> ax = fig.add_axes([0,0,1,1],frameon=False)
>>>
>>> X = np.tile(np.arange(500),(10,1)) # (10,500) shape
>>>
>>> ax.imshow(X,interpolation='nearest',aspect='auto')
>>>
>>> def go(name):
>>>
>>> for d in (1,2,3,4):
>>>
>>> w = d*5
>>> h = d
>>>
>>> fig.set_size_inches(w,h)
>>> fig.savefig("%s-%d.eps" % (name,d))
>>>
>>> rc('text', usetex=False)
>>> go("testA")
>>>
>>> rc('text', usetex=True)
>>> go("testB")
>>>
>>>
>>> ------------------------------------------------------------------------------
>>> Crystal Reports &#45; New Free Runtime and 30 Day Trial
>>> Check out the new simplified licensign option that enables unlimited
>>> royalty&#45;free distribution of the report engine for externally
>>> facing
>>> server and web deployment.
>>> http://p.sf.net/sfu/businessobjects
>>> _______________________________________________
>>> Matplotlib-devel mailing list
>>> Matplotlib-devel@lists.sourceforge.net
>>> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
>>>
>>
>

------------------------------------------------------------------------------

Register Now & Save for Velocity, the Web Performance & Operations
Conference from O'Reilly Media. Velocity features a full day of
expert-led, hands-on workshops and two days of sessions from industry
leaders in dedicated Performance & Operations tracks. Use code vel09scf
and Save an extra 15% before 5/3. http://p.sf.net/sfu/velocityconf
_______________________________________________
Matplotlib-devel mailing list
Matplotlib-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

Argg, attached a wrong patch. This one should work

         # set the paper size to the figure size if isEPSF. The
         # resulting ps file has the given size with correct bounding
         # box so that there is no need to call 'pstoeps'
        if isEPSF:
             paperWidth, paperHeight = self.figure.get_size_inches()
+ if isLandscape:
+ paperWidth, paperHeight = paperHeight, paperWidth
         else:
             temp_papertype = _get_papertype(width, height)
             if papertype=='auto':

···

On Wed, Apr 29, 2009 at 5:55 PM, Jae-Joon Lee <lee.j.joon@...149...> wrote:

Thanks Darren,

I believe that my patch only affects eps output with usetex=True,
i.e., only those "tex_*.eps" files are affected.

I found that my previous patch didn't treated the orientation of the
paper correctly.
So the bounding box of all the landscape eps outputs are incorrect.
This can be easily fixed.

Index: lib/matplotlib/backends/backend_ps.py

--- lib/matplotlib/backends/backend_ps.py (revision 7071)
+++ lib/matplotlib/backends/backend_ps.py (working copy)
@@ -1100,8 +1100,11 @@
# set the paper size to the figure size if isEPSF. The
# resulting ps file has the given size with correct bounding
# box so that there is no need to call 'pstoeps'
paperWidth, paperHeight = self.figure.get_size_inches()
+ if isLandscape:
+ paperWidth, paperHeight = paperHeight, paperWidth
else:
temp_papertype = _get_papertype(width, height)
if papertype=='auto':

Anyhow, after the landscape fix, all eps output looks fine to me.
Can you check this?

Regards,

-JJ

On Wed, Apr 29, 2009 at 3:55 PM, Darren Dale <dsdale24@...149...> wrote:

Hi Jae-Joon,

On Tue, Apr 28, 2009 at 3:49 PM, Jae-Joon Lee <lee.j.joon@...149...> wrote:

This patch is now committed to the trunk (r7068).

I think these changes have had unintended consequences. I attached a test
file I used to inspect the results when I wrote the original code that you
recently tried to improve. There are lots of combinations of large and small
figures, with usetex and without, and with no distiller, gs distiller, of
xpdf (pdftops) distiller. You have to visually inspect all of the ps/eps
files that are generated. At one time, all of these files looked acceptable,
but that was a couple years ago and I don't remember what version of
ghostscript I was using at the time (I'm sure there is a record of it
somewhere on this mailing list). WIth r7067, most of the files look fine, a
few of the large files (large paper sizes) have problems. With r7068, quite
a few examples have problems.

Darren

On Sat, Apr 25, 2009 at 8:29 AM, Ken Schutte <kts.lists@...149...> wrote:
> Yeah, that seems to work! thanks a lot,
> Ken
>
> On Fri, Apr 24, 2009 at 5:21 PM, Jae-Joon Lee <lee.j.joon@...149...> >>> > wrote:
>> ps backend, when usetex=True, uses latex with psfrag package to
>> generate the output (with some extra steps).
>> It seems that the bounding box information is not correctly recovered
>> during this process.
>> I first thought that it would be quite difficult to get this correct,
>> however the attached (relatively simple) patch seems to work fine.
>>
>> Ken, can you try the patch and see if it works?
>>
>> -JJ
>>
>>
>>
>>
>> On Thu, Apr 23, 2009 at 2:25 PM, Ken Schutte <kts.lists@...149...> >>> >> wrote:
>>> I've been trying to track down some strange behavior I was getting,
>>> and I think narrowed it down to some code that I'll paste below.
>>>
>>> I'm trying to write to .eps files, and when I have usetex=True,
>>> something is screwed up with the padding on the left, and eventually
>>> the whole image is just white.
>>>
>>> If I run this script, the 'testA-*.eps' look good, but 'testB-*' does
>>> not. The same problem happens even if I remove the ticklabels.
>>>
>>> Any tips would be appreciated.
>>> thanks,
>>> Ken
>>>
>>>
>>>
>>> ------------------------------------------------
>>> import matplotlib.pyplot as plt
>>> import numpy as np
>>> from matplotlib import rc
>>>
>>> fig = plt.figure()
>>> ax = fig.add_axes([0,0,1,1],frameon=False)
>>>
>>> X = np.tile(np.arange(500),(10,1)) # (10,500) shape
>>>
>>> ax.imshow(X,interpolation='nearest',aspect='auto')
>>>
>>> def go(name):
>>>
>>> for d in (1,2,3,4):
>>>
>>> w = d*5
>>> h = d
>>>
>>> fig.set_size_inches(w,h)
>>> fig.savefig("%s-%d.eps" % (name,d))
>>>
>>> rc('text', usetex=False)
>>> go("testA")
>>>
>>> rc('text', usetex=True)
>>> go("testB")
>>>
>>>
>>> ------------------------------------------------------------------------------
>>> Crystal Reports &#45; New Free Runtime and 30 Day Trial
>>> Check out the new simplified licensign option that enables unlimited
>>> royalty&#45;free distribution of the report engine for externally
>>> facing
>>> server and web deployment.
>>> http://p.sf.net/sfu/businessobjects
>>> _______________________________________________
>>> Matplotlib-devel mailing list
>>> Matplotlib-devel@lists.sourceforge.net
>>> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
>>>
>>
>

------------------------------------------------------------------------------

Register Now & Save for Velocity, the Web Performance & Operations
Conference from O'Reilly Media. Velocity features a full day of
expert-led, hands-on workshops and two days of sessions from industry
leaders in dedicated Performance & Operations tracks. Use code vel09scf
and Save an extra 15% before 5/3. http://p.sf.net/sfu/velocityconf
_______________________________________________
Matplotlib-devel mailing list
Matplotlib-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

I'm not sure if these are related with my patch.
Please show some example when you have time.

Regards,

-JJ

···

On Wed, Apr 29, 2009 at 4:07 PM, Ken Schutte <kts.lists@...149...> wrote:

Yes, I'm sorry, I also spoke too soon. I have found problems with
this patch - mainly things like it chopping off xlabel, titles, etc,
for some examples.

I have to make several attempts before the test script will run to completion, I get errors like:

File “/usr/lib64/python2.6/site-packages/matplotlib/dviread.py”, line 812, in find_tex_file
pipe = subprocess.Popen(cmd, stdout=subprocess.PIPE)
File “/usr/lib64/python2.6/subprocess.py”, line 595, in init
errread, errwrite)
File “/usr/lib64/python2.6/subprocess.py”, line 1084, in _execute_child
data = os.read(errpipe_read, 1048576) # Exceptions limited to 1 MB
OSError: [Errno 4] Interrupted system call

Aside from that, the results from this patch look better. Thanks,
Darren

···

On Wed, Apr 29, 2009 at 5:57 PM, Jae-Joon Lee <lee.j.joon@…149…> wrote:

Argg, attached a wrong patch. This one should work

     # set the paper size to the figure size if isEPSF. The

     # resulting ps file has the given size with correct bounding

     # box so that there is no need to call 'pstoeps'

if isEPSF:
paperWidth, paperHeight = self.figure.get_size_inches()

  •        if isLandscape:
    
  •            paperWidth, paperHeight = paperHeight, paperWidth
    
       else:
    
           temp_papertype = _get_papertype(width, height)
    
           if papertype=='auto':
    

On Wed, Apr 29, 2009 at 5:55 PM, Jae-Joon Lee <lee.j.joon@…149…> wrote:

Thanks Darren,

I believe that my patch only affects eps output with usetex=True,

i.e., only those “tex_*.eps” files are affected.

I found that my previous patch didn’t treated the orientation of the

paper correctly.

So the bounding box of all the landscape eps outputs are incorrect.

This can be easily fixed.

Index: lib/matplotlib/backends/backend_ps.py

===================================================================

— lib/matplotlib/backends/backend_ps.py (revision 7071)

+++ lib/matplotlib/backends/backend_ps.py (working copy)

@@ -1100,8 +1100,11 @@

    # set the paper size to the figure size if isEPSF. The
    # resulting ps file has the given size with correct bounding
    # box so that there is no need to call 'pstoeps'
        paperWidth, paperHeight = self.figure.get_size_inches()
  •        if isLandscape:
    
  •            paperWidth, paperHeight = paperHeight, paperWidth
    
    else:
        temp_papertype = _get_papertype(width, height)
        if papertype=='auto':

Anyhow, after the landscape fix, all eps output looks fine to me.

Can you check this?

Regards,

-JJ

On Wed, Apr 29, 2009 at 3:55 PM, Darren Dale <dsdale24@…149…> wrote:

Hi Jae-Joon,

On Tue, Apr 28, 2009 at 3:49 PM, Jae-Joon Lee <lee.j.joon@…149…> wrote:

This patch is now committed to the trunk (r7068).

I think these changes have had unintended consequences. I attached a test

file I used to inspect the results when I wrote the original code that you

recently tried to improve. There are lots of combinations of large and small

figures, with usetex and without, and with no distiller, gs distiller, of

xpdf (pdftops) distiller. You have to visually inspect all of the ps/eps

files that are generated. At one time, all of these files looked acceptable,

but that was a couple years ago and I don’t remember what version of

ghostscript I was using at the time (I’m sure there is a record of it

somewhere on this mailing list). WIth r7067, most of the files look fine, a

few of the large files (large paper sizes) have problems. With r7068, quite

a few examples have problems.

Darren

On Sat, Apr 25, 2009 at 8:29 AM, Ken Schutte <kts.lists@…149…> wrote:

Yeah, that seems to work! thanks a lot,

Ken

On Fri, Apr 24, 2009 at 5:21 PM, Jae-Joon Lee <lee.j.joon@…149…> > > >>> > wrote:

ps backend, when usetex=True, uses latex with psfrag package to

generate the output (with some extra steps).

It seems that the bounding box information is not correctly recovered

during this process.

I first thought that it would be quite difficult to get this correct,

however the attached (relatively simple) patch seems to work fine.

Ken, can you try the patch and see if it works?

-JJ

On Thu, Apr 23, 2009 at 2:25 PM, Ken Schutte <kts.lists@…149…> > > >>> >> wrote:

I’ve been trying to track down some strange behavior I was getting,

and I think narrowed it down to some code that I’ll paste below.

I’m trying to write to .eps files, and when I have usetex=True,

something is screwed up with the padding on the left, and eventually

the whole image is just white.

If I run this script, the ‘testA-.eps’ look good, but 'testB-’ does

not. The same problem happens even if I remove the ticklabels.

Any tips would be appreciated.

thanks,

Ken


import matplotlib.pyplot as plt

import numpy as np

from matplotlib import rc

fig = plt.figure()

ax = fig.add_axes([0,0,1,1],frameon=False)

X = np.tile(np.arange(500),(10,1)) # (10,500) shape

ax.imshow(X,interpolation=‘nearest’,aspect=‘auto’)

def go(name):

for d in (1,2,3,4):

   w = d*5
   h = d
   fig.set_size_inches(w,h)
   fig.savefig("%s-%d.eps" % (name,d))

rc(‘text’, usetex=False)

go(“testA”)

rc(‘text’, usetex=True)

go(“testB”)


Crystal Reports - New Free Runtime and 30 Day Trial

Check out the new simplified licensign option that enables unlimited

royalty-free distribution of the report engine for externally

facing

server and web deployment.

http://p.sf.net/sfu/businessobjects


Matplotlib-devel mailing list

Matplotlib-devel@lists.sourceforge.net

https://lists.sourceforge.net/lists/listinfo/matplotlib-devel


Register Now & Save for Velocity, the Web Performance & Operations

Conference from O’Reilly Media. Velocity features a full day of

expert-led, hands-on workshops and two days of sessions from industry

leaders in dedicated Performance & Operations tracks. Use code vel09scf

and Save an extra 15% before 5/3. http://p.sf.net/sfu/velocityconf


Matplotlib-devel mailing list

Matplotlib-devel@lists.sourceforge.net

https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

I have to make several attempts before the test script will run to
completion, I get errors like:

File "/usr/lib64/python2.6/site-packages/matplotlib/dviread.py", line 812,
in find_tex_file
pipe = subprocess.Popen(cmd, stdout=subprocess.PIPE)
File "/usr/lib64/python2.6/subprocess.py", line 595, in __init__
errread, errwrite)
File "/usr/lib64/python2.6/subprocess.py", line 1084, in _execute_child
data = os.read(errpipe_read, 1048576) # Exceptions limited to 1 MB
OSError: [Errno 4] Interrupted system call

I don't see such any error and I'm not sure if the error is related
with my patch since the ps backend does not use dviread module.

Anyhow, I just committed the patch for the landscape mode, so that
others can test.
I'll revert the changes if they seem to cause more harm than good.

Regards,

-JJ

···

On Thu, Apr 30, 2009 at 11:44 AM, Darren Dale <dsdale24@...149...> wrote:

Aside from that, the results from this patch look better. Thanks,
Darren

On Wed, Apr 29, 2009 at 5:57 PM, Jae-Joon Lee <lee.j.joon@...149...> wrote:

Argg, attached a wrong patch. This one should work

    \# set the paper size to the figure size if isEPSF\. The
    \# resulting ps file has the given size with correct bounding
    \# box so that there is no need to call &#39;pstoeps&#39;
   if isEPSF:
        paperWidth, paperHeight = self\.figure\.get\_size\_inches\(\)

+ if isLandscape:
+ paperWidth, paperHeight = paperHeight, paperWidth
else:
temp_papertype = _get_papertype(width, height)
if papertype=='auto':

On Wed, Apr 29, 2009 at 5:55 PM, Jae-Joon Lee <lee.j.joon@...149...> >> wrote:
> Thanks Darren,
>
> I believe that my patch only affects eps output with usetex=True,
> i.e., only those "tex_*.eps" files are affected.
>
> I found that my previous patch didn't treated the orientation of the
> paper correctly.
> So the bounding box of all the landscape eps outputs are incorrect.
> This can be easily fixed.
>
> Index: lib/matplotlib/backends/backend_ps.py
> ===================================================================
> --- lib/matplotlib/backends/backend_ps.py (revision 7071)
> +++ lib/matplotlib/backends/backend_ps.py (working copy)
> @@ -1100,8 +1100,11 @@
> # set the paper size to the figure size if isEPSF. The
> # resulting ps file has the given size with correct bounding
> # box so that there is no need to call 'pstoeps'
> paperWidth, paperHeight = self.figure.get_size_inches()
> + if isLandscape:
> + paperWidth, paperHeight = paperHeight, paperWidth
> else:
> temp_papertype = _get_papertype(width, height)
> if papertype=='auto':
>
>
>
>
> Anyhow, after the landscape fix, all eps output looks fine to me.
> Can you check this?
>
> Regards,
>
> -JJ
>
>
>
> On Wed, Apr 29, 2009 at 3:55 PM, Darren Dale <dsdale24@...149...> wrote:
>> Hi Jae-Joon,
>>
>> On Tue, Apr 28, 2009 at 3:49 PM, Jae-Joon Lee <lee.j.joon@...149...> >> >> wrote:
>>>
>>> This patch is now committed to the trunk (r7068).
>>
>> I think these changes have had unintended consequences. I attached a
>> test
>> file I used to inspect the results when I wrote the original code that
>> you
>> recently tried to improve. There are lots of combinations of large and
>> small
>> figures, with usetex and without, and with no distiller, gs distiller,
>> of
>> xpdf (pdftops) distiller. You have to visually inspect all of the
>> ps/eps
>> files that are generated. At one time, all of these files looked
>> acceptable,
>> but that was a couple years ago and I don't remember what version of
>> ghostscript I was using at the time (I'm sure there is a record of it
>> somewhere on this mailing list). WIth r7067, most of the files look
>> fine, a
>> few of the large files (large paper sizes) have problems. With r7068,
>> quite
>> a few examples have problems.
>>
>> Darren
>>
>>
>>>
>>> On Sat, Apr 25, 2009 at 8:29 AM, Ken Schutte <kts.lists@...149...> >> >>> wrote:
>>> > Yeah, that seems to work! thanks a lot,
>>> > Ken
>>> >
>>> > On Fri, Apr 24, 2009 at 5:21 PM, Jae-Joon Lee <lee.j.joon@...714....> >> >>> > wrote:
>>> >> ps backend, when usetex=True, uses latex with psfrag package to
>>> >> generate the output (with some extra steps).
>>> >> It seems that the bounding box information is not correctly
>>> >> recovered
>>> >> during this process.
>>> >> I first thought that it would be quite difficult to get this
>>> >> correct,
>>> >> however the attached (relatively simple) patch seems to work fine.
>>> >>
>>> >> Ken, can you try the patch and see if it works?
>>> >>
>>> >> -JJ
>>> >>
>>> >>
>>> >>
>>> >>
>>> >> On Thu, Apr 23, 2009 at 2:25 PM, Ken Schutte <kts.lists@...149...> >> >>> >> wrote:
>>> >>> I've been trying to track down some strange behavior I was
>>> >>> getting,
>>> >>> and I think narrowed it down to some code that I'll paste below.
>>> >>>
>>> >>> I'm trying to write to .eps files, and when I have usetex=True,
>>> >>> something is screwed up with the padding on the left, and
>>> >>> eventually
>>> >>> the whole image is just white.
>>> >>>
>>> >>> If I run this script, the 'testA-*.eps' look good, but 'testB-*'
>>> >>> does
>>> >>> not. The same problem happens even if I remove the ticklabels.
>>> >>>
>>> >>> Any tips would be appreciated.
>>> >>> thanks,
>>> >>> Ken
>>> >>>
>>> >>>
>>> >>>
>>> >>> ------------------------------------------------
>>> >>> import matplotlib.pyplot as plt
>>> >>> import numpy as np
>>> >>> from matplotlib import rc
>>> >>>
>>> >>> fig = plt.figure()
>>> >>> ax = fig.add_axes([0,0,1,1],frameon=False)
>>> >>>
>>> >>> X = np.tile(np.arange(500),(10,1)) # (10,500) shape
>>> >>>
>>> >>> ax.imshow(X,interpolation='nearest',aspect='auto')
>>> >>>
>>> >>> def go(name):
>>> >>>
>>> >>> for d in (1,2,3,4):
>>> >>>
>>> >>> w = d*5
>>> >>> h = d
>>> >>>
>>> >>> fig.set_size_inches(w,h)
>>> >>> fig.savefig("%s-%d.eps" % (name,d))
>>> >>>
>>> >>> rc('text', usetex=False)
>>> >>> go("testA")
>>> >>>
>>> >>> rc('text', usetex=True)
>>> >>> go("testB")
>>> >>>
>>> >>>
>>> >>>
>>> >>> ------------------------------------------------------------------------------
>>> >>> Crystal Reports &#45; New Free Runtime and 30 Day Trial
>>> >>> Check out the new simplified licensign option that enables
>>> >>> unlimited
>>> >>> royalty&#45;free distribution of the report engine for externally
>>> >>> facing
>>> >>> server and web deployment.
>>> >>> http://p.sf.net/sfu/businessobjects
>>> >>> _______________________________________________
>>> >>> Matplotlib-devel mailing list
>>> >>> Matplotlib-devel@lists.sourceforge.net
>>> >>> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
>>> >>>
>>> >>
>>> >
>>>
>>>
>>>
>>> ------------------------------------------------------------------------------
>>
>>
>>
>>>
>>> Register Now & Save for Velocity, the Web Performance & Operations
>>> Conference from O'Reilly Media. Velocity features a full day of
>>> expert-led, hands-on workshops and two days of sessions from industry
>>> leaders in dedicated Performance & Operations tracks. Use code
>>> vel09scf
>>> and Save an extra 15% before 5/3. http://p.sf.net/sfu/velocityconf
>>> _______________________________________________
>>> Matplotlib-devel mailing list
>>> Matplotlib-devel@lists.sourceforge.net
>>> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
>>
>>
>