The bundle contains a few minor cleanups that came up while I worked on
this. They are recorded in separate patches.

6 patches for repository http://darcs.net/screened:

patch 06e57cd0115de504d1b39a2f535aab85c3ed28f5
Author: Ben Franksen <ben.franksen@online.de>
Date:   Sat Sep 14 13:11:43 CEST 2019
  * remove message output in getUniqueDPatchName
  
  This function in used to derive an automatic name for a patch bundle, so we
  don't have to explain that there already exists a file with that name etc,
  since we display the final file name anyway.

patch 53dd8138ad1d47d048af85e777b7dfef940ef4c6
Author: Ben Franksen <ben.franksen@online.de>
Date:   Sat Sep 14 13:15:26 CEST 2019
  * use contextPatches when creating a patch bundle
  
  This performs an extra slightlyOptimizePatchset which can't hurt.

patch 97bd6c89444ef5d9ecee3ba685c05e785a54ffee
Author: Ben Franksen <ben.franksen@online.de>
Date:   Sat Sep 14 20:05:56 CEST 2019
  * simplify an expression in applyCmdCommon

patch 963f3b20f5081d536455954064e5c81d920d05a8
Author: Ben Franksen <ben.franksen@online.de>
Date:   Sat Sep 14 13:37:47 CEST 2019
  * fix lazy reading of inventories for obliterate with --output
  
  This one possibly fell victim to a simplification I made once, therefore the
  somewhat longer comment that explains why we have to re-read the original
  repo contents when we generate the context for the patch bundle.

patch dd705f5f9c5aa80f68306c36aff6f31f2c389be6
Author: Ben Franksen <ben.franksen@online.de>
Date:   Sat Sep 14 20:02:21 CEST 2019
  * fix lazy reading of inventories for apply command
  
  This was broken by checking availability of patches to be applied in a
  complicated and inefficient way, which as a side-effect reads all local
  inventories in our history.

patch c63580de8d04b2b2205412597b359fd74a031642
Author: Ben Franksen <ben.franksen@online.de>
Date:   Sat Sep 14 20:03:06 CEST 2019
  * extend tests/issue2293-laziness-amend.sh with more commands
  
  We now test that it works with log (--context and --last=1), amend, record,
  send, obliterate, and apply.

Attachments

• patch-preview.txt
• remove-message-output-in-getuniquedpatchname.dpatch
• unnamed

>  * remove message output in getUniqueDPatchName

I'm not sure we do always print out the final file name (e.g.
in darcs obliterate -O) but either way I agree it doesn't make
sense to say that an automatically generated file name had to be
changed.

>   * use contextPatches when creating a patch bundle

OK

>   * simplify an expression in applyCmdCommon

OK

>  * simplify an expression in applyCmdCommon

Did you swap the arguments of saveToBundle just so you could partially
apply it? I find the new order more confusing but it's no big deal.

>   * fix lazy reading of inventories for apply command

Fine

>   * extend tests/issue2293-laziness-amend.sh with more commands

Fine

>>  * remove message output in getUniqueDPatchName
> 
> I'm not sure we do always print out the final file name (e.g.
> in darcs obliterate -O)

Yes. I plan to fix that, showing the actual file name is pretty
essential for usability.

>>  * simplify an expression in applyCmdCommon
> 
> Did you swap the arguments of saveToBundle just so you could partially
> apply it? I find the new order more confusing but it's no big deal.

(This is actually in patch

  * fix lazy reading of inventories for obliterate with --output)

Yes, ahem, looks like this was the reason. I tried to re-construct why I
did it like that and I think my thinking went as follows:

- okay, result of readRepo is only ever used once
- giving a name to something that will be immediately consumed is bad
- so use a combinator, hmm, >>= seems to be the one
- oh, order of arguments doesn't fit, use flip?
- no, flip is a bit obscure; where is saveToBundle used?
- this is the only call site, let's just change its argument order

I am curious: would you consider keeping the original order of arguments
and instead using flip the better solution in this case? Or would you
rather give a name to the intermediate PatchSet?

Following up on review. Will screen immediately.

1 patch for repository http://darcs.net/screened:

patch 3bfb31206bf6299a4a806f2d35e89e7ae5fe1f1d
Author: Ben Franksen <ben.franksen@online.de>
Date:   Thu Sep 19 17:17:45 CEST 2019
  * print out name of patch bundle for obliterate --output

Attachments

• patch-preview.txt
• print-out-name-of-patch-bundle-for-obliterate-__output.dpatch
• unnamed

> I am curious: would you consider keeping the original order of arguments
> and instead using flip the better solution in this case? Or would you
> rather give a name to the intermediate PatchSet?

I think I generally find pointfree style more attractive when I'm
writing code than when I'm reading it :-) So from the perspective of a
reviewer now I'd probably say "give it a name". But I think a flip would
have been fine too. And even the actual code is not worth changing again
now.

>> I am curious: would you consider keeping the original order of arguments
>> and instead using flip the better solution in this case? Or would you
>> rather give a name to the intermediate PatchSet?
> 
> I think I generally find pointfree style more attractive when I'm
> writing code than when I'm reading it :-)

:-)) You have a certain point there.

> So from the perspective of a
> reviewer now I'd probably say "give it a name". But I think a flip would
> have been fine too. And even the actual code is not worth changing again
> now.

Agreed, I really asked (only) out of curiosity. I will try and be more
careful about pointlessly using pointfree style...